CN101214851B - Intelligent all-weather actively safety early warning system and early warning method thereof for ship running - Google Patents

Abstract

The present invention discloses an intelligent type all-weather initiative safety early-warning system of ship running and an early-warning method thereof. The early-warning system consists of a target sampling unit, a system central processing unit, a display unit, a singlechip and a safety early-warning unit. The early-warning method comprises the following steps that: a, the system is actuated; b, the target is sampled and a digital signal is processed; c, a system decision-making unit obtains information from a to be combined with the information form b to process for initiative safety early-warning decision-making analysis and calculation to obtain a final shunning proposal, and a display and an acoustic-optic warning terminal are used for the initiative safety warning synchronously.The present invention improves the sense ability of ship operating personnel to the navigation environment and the initiative safe navigation guarantee capability of the ship under the circumstance of bad visibility greatly and helps the ship operating personnel for making decision, so as to reduce the error of ship operating, improve the success rate of ship initiative shunning bump, reduce or avoid the occurrence of the traffic accidents such as ship collision, ship-against-bridge, aground, etc., and ensure the safety of the navigation transportation.

Description

Ship running intelligent all-weather actively safety forewarn system and method for early warning thereof

Technical field the present invention relates to a kind of forewarn system, particularly a kind of ship running intelligent all-weather actively safety forewarn system and method for early warning thereof.

Background technology causes the risk increase too of shipping accident generation along with the growth of China's shipping industry.Great development along with China's shipping, boats and ships develop towards maximization, high speed direction, and boats and ships quantity and waters volume of traffic and dangerous freight carrying capacity constantly increase, and marine accident happens occasionally, cause a large amount of personal casualty and economic loss, and gross pollution waters and natural environment.

By a large amount of boats and ships collision, ship are hit the bridge accident and carry out statistical results show, accident causation mainly contains three major types: first class is the human error; Second largest class is a mechanical breakdown; The third-largest class is abominable natural environment.The accident ratio that takes place under the relatively poor situation of visbility has reached more than 85%.

One of key of ship collision prevention success is to obtain the accurate information of Other dangers targets such as his ship.Use radar, the flight path of target boats and ships is determined in the change of azimuth-range that As time goes on can be by the monitored object boats and ships.But, if use radar separately, because the precision of radar in short time range is very low, the change that detects route immediately is very difficult, simultaneously, because radar will be to aerial radiation large power, electrically magnetic wave when work, thereby subject to electro countermeasure, and angle-measurement accuracy is lower.In addition, because the boats and ships that AIS or GPS much are not installed are arranged, perhaps for communication protocol interface GPS boats and ships inequality, AIS of this ship or GPS equipment can't be found.

Infrared imaging equipment has that antijamming capability is strong, and the climate environment comformability is strong, round the clock advantage such as passive detection continuously.Especially can overcome the problem that radar target is difficult to survey because of noise jamming.Along with development of science and technology, infrared imaging equipment constantly moves to maturity, and also moves towards the commercial market on the price gradually.The uncooled infrared focal plane array thermal imaging system is the main flow that small low-cost is used.From present development tendency, this market is about to occur new turnover, will become the emerging popular industry that is equal to mutually with present CCD visible image capturing technology in the period of future 5～10.Because infrared target detects, follows the tracks of and the advantage of identification, infrared imaging equipment replenishes as the outer a kind of visual information of equipment such as radar, GPS and AIS, must play a significant role in the shipping traffic system.

Infrared imaging equipment is installed on all kinds of boats and ships in key area and ocean and interior korneforos such as harbour, harbour, bridge, landing stage, dangerous river section, sluice gate, restricted area, be used for detect monitoring marine navigation, the positive effect of performance in the safe navigation of boats and ships and collision prevention.In addition, infrared image information also can merge with the information that other check implement provides, for ship's navigation provides failure-free data more.Can greatly improve the perception of behaviour's ship personnel for the navigation surrounding environment, improve the active safety navigation supportability of boats and ships under the low visibility situation, auxiliary behaviour's ship personnel make a strategic decision, reduce the error of behaviour's ship, improve initiatively collision prevention success ratio of boats and ships greatly, support personnel's the life and the safety of property reduce or avoid the generation of gross pollution waters and natural environment accident, guarantee to navigate by water transportation safety.

Summary of the invention purpose of the present invention is exactly at the deficiencies in the prior art, a kind of harbour that is installed in is provided, harbour, bridge, the landing stage, dangerous river section, sluice gate, on all kinds of boats and ships in key area and ocean and interior korneforos such as restricted area, especially large-scale pleasure-boat, passenger boat, roll-on-roll-off ship, container ship, oil carrier, the carriage of dangerous goods ship, the boats and ships of search and rescue waterborne and law enforcement, solution is under inclement weathers such as night and greasy weather rainy day, to a plurality of motion ship target on every side, bridge, bridge pier, the volume that the reef target is effectively supervised is little, function is complete, price is low, has the intelligent decision controllable function, and the civilian product of being convenient to install additional, greatly improve the perception of personnel for the navigation surrounding environment, improve the active safety navigation supportability of boats and ships under the low visibility situation, the auxilian makes a strategic decision, reduce the error of behaviour's ship, improve initiatively collision prevention success ratio of boats and ships greatly, can reduce or avoid the generation of collision prevention traffic accident in a large number, support personnel's the life and the safety of property, reduce or avoid the generation of gross pollution waters and natural environment accident, guarantee to navigate by water the ship running intelligent all-weather actively safety forewarn system and the method for early warning thereof of transportation safety.

For achieving the above object, the present invention adopts following technical scheme:

Ship running intelligent all-weather actively safety forewarn system is made of target sampling unit, system's central processing unit, display unit, micro controller system, safe early warning unit;

Described target sampling unit is made up of the outdoor intelligent high-speed The Cloud Terrace that has demoder, the thermal infrared imager that has a video frequency collection card, thermal infrared imager is fixed on the outdoor intelligent high-speed The Cloud Terrace, outdoor intelligent high-speed The Cloud Terrace connects with the system central processing unit by demoder, and thermal infrared imager connects with the system central processing unit by video frequency collection card;

Described central processing system comprises keyboard, Infrared video image digital signal processing unit, system decision-making processing unit;

Wherein,

Keyboard is used for to system decision-making processing unit input information and control command;

The Infrared video image digital signal processing unit, the data image signal that is used for the receiving target sampling unit, according to data image signal, utilize processing and analytical equipment in the memory device, calculate position, azimuth, speed and the acceleration/accel of target, and by serial communication interface result of calculation is delivered to system decision-making processing unit and use;

System decision-making processing unit is used for and will comes from the information of digital signal processing unit, information in conjunction with the input of keyboard information input unit, carry out that the active safety early warning decision is analyzed and calculate, obtain final dodge scheme after, carry out the active safety early warning simultaneously by display unit and alarm unit;

Display unit, be used to show the native system installation place around target has been carried out the image of mark, and azimuth, speed and acceleration/accel between each target and the native system installation place show the scheme of dodging, the harmful grade of recommendation simultaneously;

Micro controller system is used for the result of decision of receiving system decision-making treatment unit, the type of alarm of control alarm unit;

Alarm unit is used to receive the control signal of micro controller system, uses the sound of different frequency, the information that the light mode is represented harmful grade and dangerous direction.

System is owing to adopt integrated design, realized real-time detection to distance, orientation, speed and acceleration/accel between system installation place and a plurality of motion ship target on every side, bridge, bridge pier, the reef target, the central processing unit intellectual analysis is handled, automatically carry out danger early warning and judge decision-making, and carry out auto alarm, and finally realize the purpose of boats and ships all-weather actively safety early warning by the mode that video, literal, sound, light, electricity etc. intuitively meet human features.

Preferably, described Infrared video image digital signal processing unit is 2 floating-point signal processor TMS320C6713, the video acquisition card connection of 2 TMS320C6713 and thermal infrared imager; Described floating-point signal processor TMS320C6713 comprises video memory, programmable logic device, program store and controller, is provided with in the program store to handle and analytical equipment.It is little that TMS320C6713 has a volume, cost is low, characteristics such as stability is high, real-time is good, can realize the processing of 4 road video acquisitions simultaneously, TMS320C6713 has the alerting ability and the programmability of extremely strong processing capacity, height simultaneously, and is well positioned to meet the needs of target detect, tracking and recognizer.

Preferably, be provided with big view field image registration and splicing apparatus in the memory device in described Infrared video image digital signal processing unit, infrared sequence image is spliced into panoramic view automatically, and the panoramic view that will splice is automatically delivered to the storage of system decision-making processing unit.Big view field image registration and splicing apparatus utilize global motion estimating method, by the transformation parameter between estimated frames, have realized that the panoramic view of sequence image splices automatically.When kinematic parameter is estimated, adopted the block matching motion vector estimation of pyramid layering, effectively improve the running velocity of program, wherein add rejecting computing for abnormal mass, improve the precision of overall motion estimation greatly.In addition, under the situation that Illumination intensity between image differs.Adopt balanced equation of light computing, reached splicing effect preferably.This method can be accurately and fast the big view field image of generation video sequence, make the field range of video monitoring enlarge.The panoramic view of splicing is automatically delivered to the storage of system decision-making processing unit, can be used for post analysis evaluation and analysis on accident cause.

Preferably, described system decision-making processing unit is a PC computing machine.Adopt common PC computing machine to have low price, powerful, interface convenience, can either satisfy the requirement that the system decision-making is calculated, be convenient to again be connected with each parts of system, simultaneously because the hard-disk capacity of computing machine is big, can fully store the panoramic view that the Infrared video image of target sampling unit collection and image registration and splicing apparatus splice automatically.

Preferably, described display unit is 2 telltales, first telltale and second telltale, the image of a plurality of targets of having marked around wherein first telltale shows; Relevant Word message: distance, azimuth, speed and acceleration/accel between ship running intelligent all-weather actively safety forewarn system installation place and a plurality of on every side boats and ships, bridge, bridge pier, the reef target; The text description of the scheme of dodging that second telltale show to be recommended: comprise adopt speed change allow and/or come about allow, speed, orientation, harmful grade, dangerous direction, type of alarm.Adopt the display mode of two telltales, utilize to have obtained real-time panoramic view image, and can meet multiple goals such as boats and ships (a plurality of target), bridge (bridge pier), water surface reef and carry out actv. detection, recognition and tracking other of the boats and ships front region of travelling.By water surface zone and detected a plurality of target of splicing later panoramic view image are carried out colored painted Rendering Flags target, and simultaneously the panoramic view image that splices later panoramic view image and the colored painted Rendering Flags target of process is simultaneously displayed on the telltale, image demonstration through mark can guarantee that the user intuitively uses native system, utilize text description to satisfy user's different demands so that the user is convenient to the performance of system is estimated.

Preferably, described alarm unit is, by with external loud speaker and/or the alarm lamp of system decision-making processing unit serial communication interface bonded assembly Single-chip Controlling.The comprehensive mode of using video, sound, light, electricity etc. to meet human features is carried out auto alarm, improves the perception of behaviour's ship personnel for navigation environment, and the collision avoidance decision-making is assisted, thereby improves the ship collision prevention success ratio.

Preferably, described target sampling unit also is equipped with radar, be provided with the visible images sensor of band video frequency collection card before the radar screen, the visible images sensor connects with visible light video image digital signal processing unit in the system central processing unit by video frequency collection card, and described visible light video image digital signal processing unit is 2 floating-point signal processor TMS320C6713; Described floating-point signal processor TMS320C6713 comprises video memory, programmable logic device, program store and controller, is provided with in the program store to handle and analytical equipment; Also be provided with the information fusion unit at described central processing system, the information fusion unit, be used for the various information that come from Infrared video image digital signal processing unit and visible light video image digital signal processing unit are respectively carried out information fusion, obtain distance, azimuth, speed and the acceleration/accel of ultimate aim, simultaneously the target in the image is made a mark, and result of calculation is delivered to system decision-making processing unit by serial communication interface; Described information fusion unit is a high-speed real-time digital signal processor ADSP21060.Do not destroy former marine radar navigationsystem in this way, do not influence the use of radar, do not change electric wiring, only install visible light sensor additional on former radar, multiobject information brings great convenience in the radar in order to obtain.The inner integrated dual-port static memory device of 4Mbit of ADSP21060 has special-purpose peripheral I/O bus, and the radical function with digital information processing system is integrated on a slice chip very effectively, constitutes monolithic system easily, dwindles the volume of circuit card.High speed instruction buffer memory device (CACHE) in the sheet makes call instruction carry out with pipeline system, guarantee that every instruction finishes in the monocycle (25ns), floating-point peak value arithmetic speed is up to 120MFLOPS (million floating point operations per second), and normal floating point operation speed is 80MFLOPS.The controller of peripheral I/O bus provides six cover high-speed link mouths and two cover synchronous serial interfaces, a large amount of ADSP21060 can be constituted a loose coupling parallel processing system (PPS) with link port.In addition, treater comprises that also 3 interrupt pin, 4 sign pins, MSO-MS3 sheets select pin, make treater and other Peripheral Interfaces simple.In a word, the address space that ADSP21060 is huge, powerful addressing mode, the very long instruction word of 48bit and floating point operation ability thereof can satisfy the demand of fusion device software algorithm real-time fully.

Preferably, described ship running intelligent all-weather actively safety forewarn system is provided in a side of on bridge, harbour, harbour, landing stage, dangerous river section, restricted area, sluice gate or the ship; When this ship running intelligent all-weather actively safety forewarn system was established aboard ship, display unit also showed the image of a plurality of targets of having marked around this ship and relevant Word message: distance, azimuth, speed and acceleration/accel between this ship model, size, this ship handling parameter, load-carrying, this ship position, the real-time ship's speed of this ship, this ship and a plurality of on every side boats and ships, bridge, bridge pier, the reef target.By system being fixedly mounted on ocean and harbour, interior korneforos, harbour, bridge, the landing stage, critical areas such as restricted area, utilize Infrared video image to handle the motion boats and ships by these critical areas are carried out detection and Identification, and it is carried out track and localization handle, judge the boats and ships sense of motion in future, the operation boats and ships are monitored in real time, when boats and ships enter prewarning area, through the computing machine judgment processing, if its following sense of motion is to point to bridge pier, the restricted area, to the harbour, during situations such as the visual plant facility of harbour critical area threatens, remind the monitor staff and the marine navigator is given a warning, avoid the generation of accident by broadcasting.

Realize the method for early warning of above-mentioned ship running intelligent all-weather actively safety forewarn system, it is characterized in that may further comprise the steps:

A, start-up system

Start ship running intelligent all-weather actively safety forewarn system, by keyboard input information and control command; If ship running intelligent all-weather actively safety forewarn system is mounted on the ship, then import boats and ships model, dimension of ship, boats and ships handling parameter and the weight information of this ship by keyboard, simultaneously this ship position among this ship boat-carrying GPS and real-time ship's speed information, by serial communication interface, deliver to system decision-making processing unit;

B, target sampling and digital signal processing

Sampling of Infrared video image target and digital signal processing: drive by The Cloud Terrace with thermal infrared imager, to a/s time gap and angle step scan on every side, thereby make thermal infrared imager make a video recording to aquatic environment on every side, is digital signal through video frequency collection card with the image transitions of shooting with video-corder, and the programmable logic device of delivering in the Infrared video image digital signal processing unit carries out sequential conversion and total line control, the video memory that control row signal and graphicinformation are delivered to respectively in the digital signal processing unit is stored, video memory obtains giving after the information confirmation of programmable logic device in the Infrared video image digital signal processing unit, with processing in the program store in the Infrared video image digital signal processing unit and analytical equipment, extracting image from video memory handles and analyzes, the result who handles and analyze is delivered to the controller of Infrared video image digital signal processing unit, after controller obtains information, give acknowledgment signal of video frequency collection card, and final native system installation place and a plurality of targets on every side: boats and ships, bridge, bridge pier, azimuth between the reef, speed and acceleration information are delivered to system decision-making unit by serial communication interface;

C, system decision-making unit obtain the information from a, in conjunction with information from b, by carrying out that the active safety early warning decision is analyzed and calculate in system decision-making unit, obtain final dodge scheme after, carry out the active safety early warning simultaneously by telltale and acousto-optic alarm terminal;

Wherein:

The telltale content displayed is the image of a plurality of targets of having marked and relevant Word message around the native system installation place: distance, azimuth, speed and acceleration/accel between native system installation place and a plurality of target boats and ships, bridge, bridge pier, the reef on every side; If native system is mounted on the ship, also should show this ship model, size, this ship handling parameter, load-carrying, this ship position, the real-time ship's speed of this ship and the text description of the scheme of dodging of recommending: comprise adopt speed change allow and/or come about allow, speed, orientation;

Carry out the early warning of intelligent collision prevention active safety according to above-mentioned information, if it is dangerous, carrying out harmful grade judges, and by the sound of sound and light alarm terminal use different frequency, the information that the light mode is represented harmful grade and dangerous direction, by with system decision-making processing unit serial communication interface bonded assembly micro controller system, control external loud speaker or alarm lamp and report to the police.

Preferably, in step b, also comprise image registration and splicing, sequence image is spliced into panoramic view automatically, and the panoramic view that will splice is automatically delivered to the storage of system decision-making processing unit.Big view field image registration and splicing apparatus utilize global motion estimating method, by the transformation parameter between estimated frames, have realized that the panoramic view of sequence image splices automatically.When kinematic parameter is estimated, adopted the block matching motion vector estimation of pyramid layering, effectively improve the running velocity of program, wherein add rejecting computing for abnormal mass, improve the precision of overall motion estimation greatly.This method can be accurately and fast the big view field image of generation video sequence, make the field range of video monitoring enlarge.The panoramic view of splicing is automatically delivered to the storage of system decision-making processing unit, can be used for post analysis evaluation and analysis on accident cause.

Preferably, image processing and analysis are carried out according to following steps among the step b:

Carry out Infrared images pre-processing earlier, comprise image denoising, figure image intensifying and sharpening, image rectification, movement background correction; Infrared image is a set of being disturbed to form by real scene image, imaging noise and imaging.Suppose f (x, the y) infrared image that obtains of expression imaging system then include the target infrared scene image and can retouch speed and be:

f(x，y)＝f _T(x，y)+f _B(x，y)+n(x，y)+n ₁(x，y)

f _T(x y) is target gray value; f _B(x y) is background image, and (x y) is expressed as picture noise, n to n ₁(x y) is expressed as picture and disturbs.Background image f _B(x y) has long persistence length usually, and it has occupied scene image f (x, y) low-frequency information in the spatial frequency.Simultaneously, since the irregularity of scene and sensor internal heat distribution, background image f _B(x is a non-stationary process y), and local gray-value may have bigger variation in the image, in addition, and f _B(x y) also comprises high fdrequency component in the segment space frequency domain, and they mainly are distributed in the edge in each homogeneity district of Background image.

(x y) introduces in imaging process imaging noise n, is some microvariations that are superimposed upon on the infrared image random site; And n is disturbed in imaging ₁(x, y) then be since photosensitive unit response heterogeneity of infrared imaging or dummy argument cause it infrared image at random or form some wrong image data point on the fixed position.Therefore, n is disturbed in imaging ₁(x, y) in image, show as the pixel gray value much larger than or less than some isolated points of neighborhood intermediate value around it, the purpose of infrared image denoising is exactly to estimate the real scene image from image f.

Movement background is proofreaied and correct because imageing sensor is installed on the motion platform sometimes, even or be installed on the static platform, also may be because some interference causes the sensor shake, the shake of sensor will cause background shake.By the global motion parameter estimation techniques.It at first estimates the kinematic parameter of background, utilizes the kinematic parameter that estimates to do background correction then, and multiple image is proofreaied and correct under the same coordinate system.

Directly, extract bridge, bridge pier and reef feature then, carry out actv. and follow the tracks of and discern image region segmentation;

Adopting two kinds of algorithms of different combinations to carry out sea horizon respectively extracts, wherein, the steps in sequence that the extraction algorithm combination that article one may sea horizon comprises is: the image iteration threshold is cut apart, Roberts gradient operator rim detection, refinement, utilize the Hough conversion to extract article one sea horizon; Second may sea horizon the steps in sequence that comprises of extraction algorithm combination be: Roberts gradient operator rim detection and binaryzation, refinement, utilize the Hough conversion to extract the second sea horizon; Get in two possible sea horizons a sea horizon near the image lower end as the final sea horizon that extracts, and sea horizon carried out confidence level judge, then sea horizon up and down the image-region in the certain limit as region of interest ROI;

Based on the position relation of sea horizon and ship target, the ship target image generally is in the sea horizon zone.This is determined by the middle distance planar imaging.Target can not break away from sea horizon fully and be in a day dummy section, also can not break away from sea horizon fully and is in other zones such as land, valley.Therefore the correct sea horizon that detects, then sea horizon up and down the image-region in the certain limit as area-of-interest (ROI), the detection of follow-up infrared ship target, tracking and identification, just dwindled the image processing scope greatly, and greatly avoided the interference of some high-radiation areas on aerial cloud, ripples, land or the valley, thereby make that the calculated amount of various algorithms is significantly reduced, guaranteed the requirement of algorithm real-time.

The sea horizon leaching process that the present invention adopts is: at first original image is carried out picture quality and estimate, determine whether image is carried out pretreatment according to evaluation result, carrying out sea horizon with two kinds of algorithms of different combinations simultaneously then extracts, wherein, the steps in sequence that the extraction algorithm combination that article one may sea horizon comprises is: the image iteration threshold is cut apart, Roberts gradient operator rim detection, refinement, utilize the Hough conversion to extract article one; Second may sea horizon the steps in sequence that comprises of extraction algorithm combination be: Roberts gradient operator rim detection and binaryzation, refinement, utilize the Hough conversion to extract the second sea horizon.After finishing two possible sea horizons extractions, get a final sea horizon that extracts of sea horizon conduct near the image lower end.

Image quality evaluation: adopt error of mean square (MSE) to come picture quality is estimated, be used for determining whether image is carried out pretreatment.

If MSE＞K then carries out the image pretreatment

If MSE≤K then does not carry out the image pretreatment

Get K=25.

First order image pretreatment: in the sky and ground region in infrared image more than the sea horizon, the existence of bridge, surface structures, continuous targets such as rock is arranged, the gray value of these targets generally all is in the high grade grey level of image, gradient between they and the surrounding environment is often greater than the gradient of sea horizon both sides, when especially these targets present the continuous linearity of nearly horizontal direction and distribute in image, cause present existed algorithms all can not correctly extract sea horizon.Adopting following method to eliminate the high target of these brightness values disturbs: the gray value of supposing pixel in the image is that (x, y), pixel adds up to M * N to f in the image, need the high pixel ratio of elimination brightness value be R, f _{M * N * R}(x, the y) gray value of the M * N * R pixel of expression when gray value is arranged from high to low, r represent pixel (x, eight lowest gray value in pixel y), behind the pretreatment pixel corresponding gray be g (x, y), then

if?f(x，y)≥f _M×N×R(x，y)then?g(x，y)＝r

if?f(x，y)＜f _M×N×R(x，y)then?g(x，y)＝f(x，y)

Owing to only reduce the brightness value of part high luminance values pixel, can the accurate extraction of sea horizon not impacted in this step.R gets 0.05.

Second stage image pretreatment: the water surface zone in infrared image below the sea horizon, the interference of strong ripples is to cause sea horizon can not extract another principal element.Strong ripples disturb at the shared pixel in water surface zone more, its grey value profile is near the average of entire image, and adopt following method to remove strong ripples and disturb: calculate image average fmean, the pixel corresponding gray is h (x behind the pretreatment of the second stage, y), then

if?f(x，y)＞fmean?then?h(x，y)＝f(x，y)-fmean

if?f(x，y)≤fmean?then?h(x，y)＝0

Behind the pretreatment of the second stage, most strong ripples disturb and are suppressed.Same this step can not impact the accurate extraction of sea horizon owing to only reduce the brightness value that the strong ripples of part disturb pixel.

The image iteration threshold is cut apart: entire image is regarded as by two zones and is constituted in the water transport environment, i.e. following water surface zone and sky and the ground region more than the sea horizon of sea horizon.When the MSE of original image≤K because the interference that is subjected to is little, directly image is carried out iteration threshold and cuts apart, can realize to two zones carry out actv., failure-free is cut apart; When the MSE of original image＞K, behind the image pretreatment, removed most of interference, at this moment again image is carried out iteration threshold and cuts apart, can realize to two zones carry out actv., failure-free is cut apart.

Roberts gradient operator rim detection: the Roberts gradient of asking for each picture element in the image respectively.

Binaryzation: gradient image is adopted edge threshold strategy binaryzation.In piece image, non-number of edge points is occupied certain proportion in total image pixel is counted out, and cooresponding factor of proportionality is expressed as Ratio.According to the corresponding histogram of image gradient value, begin progressively accumulated image point from low Grad grade, when the number that adds up reached image total pixel number purpose Ratio, cooresponding image gradient value was segmentation threshold.Ratio gets 0.95.

Refinement: have pixel in the bianry image and join together, influence the extraction precision and the real-time of Hough conversion, so will carry out refinement.The principle of refinement is for vertically being reduced to line segment a pixel.

Utilize the Hough conversion to extract sea horizon:

Utilize the Hough conversion to extract straight line, realized a kind ofly from the mapping of image space to parameter space, its basic thought is the duality of point one line.Polar equation by straight line: ρ=xcos θ+ysin θ carries out the Hough conversion, promptly uses the point on the cathetus of sine curve presentation video space.Herein parameter space is dispersed and change into an accumulator/accum battle array, with the every bit (x in the image, y) be mapped in the cooresponding a series of accumulator/accums of parameter space, make cooresponding accumulator value add 1, if comprise straight line in the image space, then in parameter space, there is a cooresponding accumulator/accum local maximum can occur.By detecting this local maximum, can determine that (ρ θ), thereby detects straight line with the cooresponding a pair of parameter of this straight line.

The reliability of sea horizon is the probabilistic problem of minimizing that a plurality of evidences of need are supported the same fact, with evidence theory the sea horizon that extracts is carried out confidence level for this reason and judges.

(1) whether sea horizon is positioned at possible sea horizon zone;

After this identical installation, must being in the sub regions of sea horizon in subregion, then extracted the confidence level M1=1 of sea horizon as the sea horizon that extracts, otherwise, extract the confidence level M1=0 of sea horizon.Pass through formula:

M1＝1，y1≤y≤y2

M1=0, y＜y1 or y＞y2

In the formula: y1, y2 are respectively the intermediate image vegetarian refreshments row-coordinate of the highest, minimum possibility sea horizon; Y is for extracting the intermediate image vegetarian refreshments row-coordinate of sea horizon.

(2) judge by the contrast ratio of sea horizon upper and lower subregion, obtain the confidence level of sea horizon;

Because the gray scale of sea horizon upper area generally will be higher than lower zone, when upper and lower region contrast contrast is big more, then the confidence level of sea horizon is high more.Calculate by formula:

${\mathrm{<figure-callout\; id="2"\; label="M"\; filenames="S2008100692312E00011.png,S2008100692312E00021.png"\; state="\{\{state\}\}">M</figure-callout>}}_2=\frac{[\underset{{\mathrm{<figure-callout\; id="1"\; label="h"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">h</figure-callout>}}_1<x<{\mathrm{<figure-callout\; id="2"\; label="h"\; filenames="S2008100692312E00011.png,S2008100692312E00021.png"\; state="\{\{state\}\}">h</figure-callout>}}_2}{\mathrm{\&Sigma;}}\underset{{\mathrm{<figure-callout\; id="1"\; label="w"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">w</figure-callout>}}_1<y<w_2}{\mathrm{\&Sigma;}}f(x,y)-\underset{{\mathrm{<figure-callout\; id="1"\; label="h"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">h</figure-callout>}}_1+\mathrm{\&Delta;h}<x<{\mathrm{<figure-callout\; id="2"\; label="h"\; filenames="S2008100692312E00011.png,S2008100692312E00021.png"\; state="\{\{state\}\}">h</figure-callout>}}_2+\mathrm{\&Delta;h}}{\mathrm{\&Sigma;}}\underset{{\mathrm{<figure-callout\; id="1"\; label="w"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">w</figure-callout>}}_1<y<w_2}{\mathrm{\&Sigma;}}f(x,y)]}{\underset{{\mathrm{<figure-callout\; id="1"\; label="h"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">h</figure-callout>}}_1<x<{\mathrm{<figure-callout\; id="2"\; label="h"\; filenames="S2008100692312E00011.png,S2008100692312E00021.png"\; state="\{\{state\}\}">h</figure-callout>}}_2}{\mathrm{\&Sigma;}}\underset{{\mathrm{<figure-callout\; id="1"\; label="w"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">w</figure-callout>}}_1<y<w_2}{\mathrm{\&Sigma;}}f(x,y)}$

The be associated confidence level M3 of evidence of the intermediate image vegetarian refreshments of the sea horizon that frame extracts (3).

${\mathrm{<figure-callout\; id="3"\; label="M"\; filenames="S2008100692312E00011.png"\; state="\{\{state\}\}">M</figure-callout>}}_3=\frac{1}{|y\left(t\right)-y(t-1)|},$ During y (t)=y (t-1), M ₃=1

Y (t), y (t-1) are respectively the intermediate image vegetarian refreshments row-coordinate of the sea horizon that extracts by present frame and former frame image, and when the difference of the two is big more, then the confidence level of sea horizon is low more.

(4) utilize the composition rule of D-S, obtain the comprehensive confidence level M of sea horizon, M=M is arranged ₁M ₂M ₃

Make that Mt is the thresholding of a suitable selection, when M＞Mt, think that the extraction of sea horizon of this two field picture is correct; Otherwise above-mentioned criterion can't be confirmed the correctness of the sea horizon of this frame.At this moment, preserve position, the regional up and down contrast ratio of sea horizon of former frame sea horizon.Proceeding the identification of the sea horizon of next frame handles.

Correct detect behind the sea horizon sea horizon up and down the interior image-region of certain limit as area-of-interest (ROI).

Interesting image regions ROI is carried out the evaluation of picture quality, when evaluation result is 1, adopt infrared target detection algorithm based on single-frame images; When evaluation result is 0, adopt infrared target detection algorithm based on image sequence;

Use the two-dimensional wavelet transformation in the wavelet analysis method to realize frequency selection and multiple dimensioned decomposition respectively, suppress background noise and strengthen target; Original image is carried out separating of low frequency part and HFS, respectively each low frequency and high fdrequency component are carried out multiresolution analysis then, extract target signature, carry out target detect;

In the fractal method according to the fractal characteristic of man-made targets such as boats and ships and bridge pier with the dimensional variation characteristics more violent than natural background, extract the Multi-scale Fractal image by the ROI subimage that carries out through the fuzzy filter method after the figure image intensifying; At last, utilize probabilistic relaxation that Multi-scale Fractal is carried out target detect; Fractal model be a kind of be suitable for describing have complicated and irregularly shaped mathematics model, its groundwork is to utilize natural scene and the man-made target difference on fractal dimension, the line style of logarithm value is estimated between the metric that common fractal dimension is each yardstick summer of calculating and the scale of measurement, think in each range scale that promptly fractal dimension is a constant value, this meets the feature of desirable fractal model.But for most of natural scenes, just in certain range scale, be similar to and have fractal characteristic, and real image is influenced by imaging noise, quantizing error etc., natural scene can not be described with the standard fractal dimension under many circumstances, and therefore the difference from the standard fractal dimension is distinguished natural background and man-made target often can not be obtained desirable effect.Consider that target detect is the fractal characteristic difference of utilizing between target and the background, do not need its fractal parameter of accurate Calculation, and background is different with the difference of dimensional variation with target in the image, utilizes the difference of both multiple dimensioned variations to carry out target detect.

Earlier the ROI subgraph is carried out medium filtering in the Mathematical Morphology Method, the pixel of getting brightness value maximum in the filtered image then image that serves as a mark; Original image is through top cap conversion, and the image after the row iteration threshold value of going forward side by side is cut apart carries out morphology reconstruct as mask images, realizes that infrared ship target detects; Morphology reconstruct: the thought of reconstruct is approached mask images by marking image is constantly expanded exactly, thereby recovers certain a part of or whole mask images of mask images, and this depends on choosing of marking image.The characteristics of reconstruct are by the choosing of marking image, and can extract own interesting areas in the mask images.

Define by following iterative process by f (marking image) reconstruct g (mask images):

H1 is initialized as marking image f;

Create 3x3 structural element B, each element all is 1 among the B, limits together with rule;

Repeat h _K+1=(h _k⊕ B) ∩ g is up to h _K+1=h _k

Attention: mark f must be the subclass of g, and the image after the reconstruct is the subclass of mask images g.

Use the evidential reasoning combination at last the target detect result from distinct methods is carried out the evidential reasoning combination, adopt Dempster evidence compositional rule that evidence is synthetic, obtain a high infrared ship target testing result of confidence level identification; Evidence theory can not need prior probability and conditional probability density, adopt belief function rather than probability as tolerance, " interval estimation " adopted in description to uncertain information, rather than the method for " point estimation ", differentiation do not know with uncertain aspect and reflect that accurately the evidence-gathering aspect demonstrates very big alerting ability.

Preferably, image processing and analysis are further comprising the steps of among the step b:

Using artificial neural networks technology, Fuzzy Inference realize the moving infrared ship target tracking of intelligent multimachine; Using artificial neural networks carries out the multimachine tracking of maneuvering target, has good adaptive, self-organized learning and association and fault-tolerant ability, by study and training, stronger judgement and discernibility is arranged, and can independently find the way of dealing with problems; Use Fuzzy Inference to a plurality of targets of being followed the tracks of assessment that impends, judgement target type, target velocity, thus infer their threat level; Artificial neural net (ANN) is compared with traditional adaptive system at random, by study and training, stronger judgement and discernibility is arranged, and can independently find the way of dealing with problems.And its fault-tolerant ability is very strong, under data and environment uncertain condition, especially under the situation of much noise and interference, still can carry out target detect, parameter estimation, target's feature-extraction and identification, system modelling etc.Fuzzy reasoning is to have imitated the reasoning process of human thinking to real things; the multiple target tracking process " may be related " or " may be not related " between the target occur through regular meeting, and target " may belong to a class target " or analogues such as " may not belong to a class target ".The utilization fuzzy reasoning is judged target type, target velocity etc. to a plurality of targets of being followed the tracks of assessment that impends, thereby infers their threat level, can accurately propose the collision avoidance scheme for grasping the ship personnel, provides a favorable security.The traditional method of these utilizations is difficult to satisfy the demand on the engineering, and the utilization fuzzy reasoning can simplify a problem, to satisfy the needs of real-time in the target tracking.

Preferably, image processing and analysis are further comprising the steps of among the step b:

Employing increases self adaptation and learning ability in the classical mode recognizer, utilize the image context, the knowledge base that incorporates artificial intelligence technology carries out the identification of infrared ship target, carry out the extraction of target signature in cut zone, difference extracting position feature, shape facility, size characteristic, radiation feature and the feature of extracting based on wavelet analysis, and, carry out the identification of infrared ship target the input end of these features as the RBF neural network.Design-calculated RBF neural network of the present invention has the three-layer network structure, and input layer comprises 8 input parameters, and hidden layer has 12 nodes, the type of output layer input target.

Preferably, further comprising the steps of among the step b:

B1, sampling of visible light video image target and digital signal processing: with being installed in radar screen fwd visible images sensor, the radar screen is made a video recording, obtain radar screen picture, is digital signal through video frequency collection card with the image transitions of shooting with video-corder, and the programmable logic device of delivering in the visible light video image digital signal processing unit carries out sequential conversion and total line control, the video memory that control row signal and graphicinformation are delivered to respectively in the visible light video image digital signal processing unit is stored, video memory obtains giving after the information confirmation of programmable logic device in the visible light video image digital signal processing unit, with processing in the program store in the visible light video image digital signal processing unit and analytical equipment, extracting image from video memory handles and analyzes, the result who handles and analyze is delivered to the controller of visible light video image digital signal processing unit, after controller obtains information, give acknowledgment signal of video frequency collection card, and final native system installation place and a plurality of targets on every side: boats and ships, bridge, bridge pier, distance between the reef, the azimuth, speed and acceleration information, the information fusion unit of delivering in the central processing system by serial communication interface carries out the breath fusion;

B2, information fusion: the information fusion unit obtains coming from the information of Infrared video image digital signal processing unit and visible light video image digital signal processing unit, give Infrared video image digital signal processing unit and confirmation of visible light video image digital signal processing unit among the corresponding step b respectively, the information fusion unit will the Infrared video image digital signal processing unit obtains from step b native system installation place and a plurality of target boats and ships on every side, bridge, bridge pier, azimuth between the reef, speed and acceleration/accel, and the native system installation place that obtains of visible light video image digital signal processing unit and a plurality of target boats and ships on every side, bridge, distance between the reef, the azimuth, speed and acceleration/accel, handle through information fusion, obtain final fusion results information, be the native system installation place and a plurality of target boats and ships on every side, bridge, bridge pier, distance between the reef, the azimuth, the accurate information of speed and acceleration/accel, simultaneously the target in the image is made a mark, deliver to system decision-making unit by serial communication interface.

Preferably, according to the information of importing among the step a, distance, azimuth, speed and acceleration/accel between the native system installation place that obtains among the integrating step b and a plurality of on every side target boats and ships, bridge, bridge pier, the reef; Adopt the fuzzy expert system method,, comprehensively become the principal component of separate quadrature, finally determine the ship collision prevention degree of risk, carry out the active early warning decision based on this each early warning decision key element and influence degree thereof; If native system is mounted on the ship, according to this ship model that obtains among the step a, size, this ship handling parameter, load-carrying, this ship position, the real-time ship's speed of this ship, distance, azimuth, speed and acceleration/accel between the native system installation place that obtains among the integrating step b and a plurality of on every side target boats and ships, bridge, bridge pier, the reef; Adopt the fuzzy expert system method,, comprehensively become the principal component of separate quadrature, finally determine the ship collision prevention degree of risk, carry out the active early warning decision based on this each early warning decision key element and influence degree thereof.

The active safety forewarn system is by complication system that people-ship-the environment three elements are formed.Thereby the factor that influences people, boats and ships, environment three elements produces significant effects to the active safety early warning decision.On the basis of the collision prevention information of utilizing marine navigation equipment to obtain, intend adopting the fuzzy expert system method to determine the ship collision prevention degree of risk, carry out the active early warning decision based on this.

Boats and ships under steam, it is x that each sensor obtains the early warning decision key element ₁, x ₂, x ₃X _n, each key element is different to the influence degree of early warning decision, and also interacts between the key element.

If i key element is w to the influence degree of early warning decision _i(1≤i≤n), n sample establishing extraction is X:x ₁ ^t, x ₂ ^t..., x _n ^t((1≤t≤n)). n characteristic index comprehensively becomes independently principal component y of n mutually orthogonal now ₁, y ₂..., y _n, write as matrix form and be

Y＝C·X (2)

Wherein

$C=\left[\begin{array}{ccc}{\mathrm{<figure-callout\; id="11"\; label="C"\; filenames="S2008100692312E00011.png"\; state="\{\{state\}\}">C</figure-callout>}}_{11}& ...& C_{1n}\\ ...& ...& ...\\ {\mathrm{<figure-callout\; id="1"\; label="C\; n"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">C</figure-callout>}}_n& ...& C_{\mathrm{nn}}\end{array}\right],$ $Y=\left[\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="y"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">y</figure-callout>}}_1\\ \&CenterDot;\&CenterDot;\&CenterDot;\\ y_n\end{array}\right]$

Fuzzy expert system can be handled probabilistic data and proposition, promptly can be in [0,1] middle value, and its adopts fuzzy technology means such as fuzzy set, fuzzy number and fuzzy relation to represent and handle the uncertainty and the inexactness of knowledge.Because in the ship running process, uncertain factor is very many, so the active safety early warning decision adopts the practicality of fuzzy expert system energy enhanced system.

The active safety forewarn system is made up of three main portions such as information collection, data handing, decision-makings.Preferably, the a plurality of target boats and ships in this ship running intelligent all-weather actively safety forewarn system that obtains according to the Infrared video image digital signal processing unit among the step b among the step b2 and the place ahead, bridge, bridge pier, azimuth between the reef, speed and acceleration/accel, the a plurality of target boats and ships of this ship running intelligent all-weather actively safety forewarn system and the place ahead that obtain with visible light video image digital signal processing unit from step b1, bridge, distance between the reef, the azimuth, speed and acceleration/accel, utilize high-precision measurement of angle of thermal infrared imager and the high-precision range observation of radar, utilize message complementary sense, by information fusion technology, provide accurate estimation to the target location; The employing centralized processing method that merges characteristic layer realizes that radar and thermal infrared imager merge, at first extract the target barycenter in the infrared image, with the least-squares estimation method the redundant angle technology of thermal infrared imager is compressed then, to produce in time the pseudo-measurement of angle of aiming at radar surveying, carry out fusion treatment with the measurement of azimuth of radar respectively again, merge estimation to obtain synchrodata, at last the data that obtain based on radar and infrared fusion are used to upgrade the dbjective state of filter, the fusion process of decision-making level adopts distributed approach, at first by radar and the infrared flight path of setting up separately about target, and then carry out radar and the related of infrared flight path and merge.

Radar is as active sensor, because can round-the-clock measurement and provide target complete location information, thereby bringing into play important effect aspect target acquisition and the tracking.But because radar will be to aerial radiation large power, electrically magnetic wave in when work, thereby subjects to electro countermeasure, and angle-measurement accuracy is lower.

Thermal infrared imager is not to any energy of aerial radiation, and it is surveyed by the heat energy of receiving target radiation and locatees, and has stronger antijamming capability, and simultaneously thermal infrared imager also has advantages such as the high and target recognition ability of angle accuracy is strong, but can not find range from.

Utilize high-precision range observation of radar and the high-precision measurement of angle of thermal infrared imager, utilize message complementary sense,, can provide accurate estimation, improve tracking and identification target to the target location by information fusion technology.

(1) sensor measurement model

What the infrared imaging thermal imaging system was measured is the azimuth and the angular altitude at object brightness center, and the brightness center of hypothetical target overlaps with barycenter here, and then its measurement model is:

${\mathrm{\&theta;}}_I\left(k\right)=\mathrm{\&theta;}\left(k\right)+{\mathrm{\&upsi;}}_{{\mathrm{\&theta;}}_I}\left(k\right),$

${\mathrm{\&phi;}}_I\left(k\right)=\mathrm{\&phi;}\left(k\right)+{\mathrm{\&upsi;}}_{{\mathrm{\&phi;}}_I}\left(k\right).$

θ in the formula _I(k), φ _I(k) be infrared observed reading, θ (k), φ (k) are actual angle,

Be the measurement of angle noise, their average is zero, variance is a Gaussian white noise.The dbjective state vector is elected position, speed and the acceleration/accel in the inertial system as, promptly

$X\left(k\right)={\left[x\stackrel{\&CenterDot;}{\left(k\right)\stackrel{\&CenterDot;}{x}}\left(k\right)\stackrel{\&CenterDot;\&CenterDot;}{x}\left(k\right)y\left(k\right)\stackrel{\&CenterDot;}{y}\left(k\right)\stackrel{\&CenterDot;\&CenterDot;}{y}\left(k\right)z\left(k\right)\stackrel{\&CenterDot;}{z}\left(k\right)\stackrel{\&CenterDot;\&CenterDot;}{z}\left(k\right)\right]}^T,$ Then have

$\left[\begin{array}{c}{\mathrm{\&theta;}}_I\left(k\right)\\ {\mathrm{\&phi;}}_I\left(k\right)\end{array}\right]=\left[\begin{array}{c}\arctan [z\left(k\right)/x\left(k\right)]\\ \arctan [y\left(k\right)/\sqrt{x^2\left(k\right)+z^2\left(k\right)}]\end{array}\right]+\left[\begin{array}{c}{\mathrm{\&upsi;}}_{{\mathrm{\&theta;}}_I}\left(k\right)\\ {\mathrm{\&upsi;}}_{{\mathrm{\&phi;}}_I}\left(k\right)\end{array}\right].$

Radar is the distance and the azimuth of measurement target directly, and its measurement model is

$r_R\left(k\right)=r\left(k\right)+{\mathrm{\&upsi;}}_{r_R}\left(k\right),$

${\mathrm{\&theta;}}_R\left(k\right)=\mathrm{\&theta;}\left(k\right)+{\mathrm{\&upsi;}}_{{\mathrm{\&theta;}}_R}\left(k\right),$

${\mathrm{\&phi;}}_R\left(k\right)=\mathrm{\&phi;}\left(k\right)+{\mathrm{\&upsi;}}_{{\mathrm{\&phi;}}_R}\left(k\right).$

R in the formula _R(k), θ _R(k), φ _R(k) be the observed reading of radar, r (k), θ (k), φ (k) are actual value,

Be to measure noise, their average is zero, variance is a Gaussian white noise.The dbjective state vector is chosen the same, then has

$\left[\begin{array}{c}{r}_R\left(k\right)\\ {\mathrm{\&theta;}}_R\left(k\right)\\ {\mathrm{\&phi;}}_R\left(k\right)\end{array}\right]=\left[\begin{array}{c}\sqrt{x^2\left(k\right)+y^2\left(k\right)+z^2\left(k\right)}\\ \arctan [z\left(k\right)/x\left(k\right)]\\ \arctan [y\left(k\right)/\sqrt{x^2\left(k\right)+z^2\left(k\right)}]\end{array}\right]+\left[\begin{array}{c}{\mathrm{\&upsi;}}_{r_R}\left(k\right)\\ {\mathrm{\&upsi;}}_{{\mathrm{\&theta;}}_R}\left(k\right)\\ {\mathrm{\&upsi;}}_{{\mathrm{\&phi;}}_R}\left(k\right)\end{array}\right].$

(2) structure mode and the algorithm of radar, infrared image signal fusion and decision system

What characteristic layer merged act as: utilize the characteristic information of IR imaging target identification and tracking to help the target recognition and the tracking of radar, thereby improve the target detection probability of system and reduce false-alarm probability; What decision-making level merged act as: when target relative distance is far away, guide the servo control unit tracking target of thermal infrared imager according to the tracking decision information of radar module, target is dropped in the visual angle of thermal infrared imager, thermal infrared imager can be discerned and tracking target voluntarily by imaging analysis during near target with box lunch, thereby can remedy the near deficiency of infrared imaging thermal imaging system ranging coverage, performance infrared imaging thermal imaging system is followed the tracks of the high advantage of precision of decision information near target the time.When losing tracking target ability or tracking target ability because of reason one sensor assemblies such as interference, can correct the target tracking ability of this sensor assembly that is disturbed according to the tracking decision information of another sensor assembly, thereby improve the anti-interference resistance of system, improved the reliability of whole target recognition track channel in addition, in case because of software or the wherein a certain sensor of hardware fault have lost target recognition and traceability, the fusion center decision controller still can and be followed the tracks of the correct tracking target of decision signal according to the target recognition of another sensor.

Because the data collection rate of thermal infrared imager is apparently higher than radar, therefore, characteristic layer merges the centralized processing method that adopts, realize that the basic thought that radar and thermal infrared imager merge is: at first extract the target barycenter in the infrared image, with the least-squares estimation method the redundant angle technology of thermal infrared imager is compressed then, to produce in time the pseudo-measurement of angle of aiming at radar surveying, carry out fusion treatment with the measurement of azimuth of radar respectively again, merge to estimate to obtain synchrodata, at last the data that obtain based on radar and infrared fusion are used to upgrade the dbjective state of filter.

Merge radar and infrared data fusion for decision-making level, adopt distributed approach,, and then carry out radar and the related of infrared flight path and fusion promptly at first by radar and the infrared flight path of setting up separately about target.

The invention has the advantages that:

(1) system is owing to adopt integrated design, realized real-time detection to distance, orientation, speed and acceleration/accel between a plurality of targets in travel boats and ships and the place ahead, central processing unit (computing machine) intellectual analysis is handled, automatically carry out danger early warning and judge decision-making, and carry out auto alarm, and finally realize the purpose of boats and ships all-weather actively safety early warning by the mode that video, literal, sound, light, electricity etc. intuitively meet human features.

(2) utilization Eltec, bus of circuit technology and principle of design, central processing unit has adopted optimizes integrated design, signal is modulated, input buffering, sampling maintenance, data acquisition, A/D conversion, CPU processing, D/A conversion, signal amplification, Display and Alarm Circuit and executive control system circuit are through optimal combination, integrated, form function in the central processing unit of one, it is integrated to have realized that data sampling, conversion and intellectual analysis are handled and controlled.

(3) infrared thermal imaging technique is applied to the safe navigation and the collision prevention of boats and ships, is not subjected to electromagnetic interference effect.

(4) by the information fusion single-purpose computer to merging from shipborne radar, infrared information respectively, improved system target detection probability, reduce false-alarm probability; The anti-interference resistance of raising system, improved the reliability of whole target recognition track channel.

(5) but system's all weather operation, especially in night, greasy weather and rainy day of low visibility;

(6) variation of type of alarm, the intellectuality of active safety early warning decision, it is accurate, timely, abundant, convenient, directly perceived that behaviour's ship personnel obtain multiple-object information;

(7) volume is little, function is complete, price is low;

(8) system has stability, real-time and accuracy.About the 3 kilometers bridges far away of native system of adjusting the distance: rate of false alarm≤1%; Rate of failing to report≤1%; Detection, tracking and recognition correct rate 〉=98%.About the 1 kilometer boats and ships far away of native system of adjusting the distance: rate of false alarm≤2%; Rate of failing to report≤2%; Detection, tracking and recognition correct rate 〉=98%.

(9) installation of each parts of system, because the principle of having taked not destroy former boats and ships external form, not changed electric wiring, only on former boats and ships basis, install additional, for all kinds of boats and ships installation system parts bring great convenience.

Ship running intelligent all-weather actively safety method for early warning and forewarn system thereof, can greatly improve the perception of behaviour's ship personnel for the navigation surrounding environment, improve the active safety navigation supportability of boats and ships under the low visibility situation, auxiliary behaviour's ship personnel make a strategic decision, reduce the error of behaviour's ship, improve initiatively collision prevention success ratio of boats and ships greatly, can reduce or avoid boats and ships to bump against in a large number, ship hits bridge, the generation of traffic accident such as hit a submerged reef, support personnel's the life and the safety of property, reduce or avoid the generation of gross pollution waters and natural environment accident, guarantee to navigate by water transportation safety

Description of drawings Fig. 1 is a structural representation of the present invention;

Fig. 2 is the structural representation of Infrared video image digital signal processing unit of the present invention;

Fig. 3 is the step block diagram of infrared target sampling of the present invention, processing, recognition methods;

Fig. 4 is the step block diagram of safe early warning side of the present invention;

The present invention is further illustrated below in conjunction with the specific embodiment for the specific embodiment, but therefore do not limit the present invention among the described scope of embodiments:

Embodiment 1: system is installed on the bridge

Ship running intelligent all-weather actively safety forewarn system is fixedly mounted on the bridge, be used to realize active safety early warning to the pier anticollision of this bridge, utilize Infrared video image to handle the motion boats and ships by bridge are carried out detection and Identification, and it is carried out track and localization handle, judge the boats and ships sense of motion in future, the operation boats and ships are monitored in real time, when boats and ships enter prewarning area, the size of prewarning area determines according to actual conditions, through the computing machine judgment processing, when the sense of motion of boats and ships was not the sensing bridge pier, safety of ship travelled, and does not start warning device; If the motion vector direction of boats and ships is to point to bridge pier, start warning device, give the alarm by broadcasting, remind the deck officer to take to change course or speed change etc. dodged measure, guarantees all safety of bridge pier and boats and ships.

Certainly, according to the difference of ship running intelligent all-weather actively safety forewarn system fixed installation position and field condition, also can be fixedly mounted on system on the harbor traffic safety management monitor console near harbour and the harbour the land building, on the bridge, the relevant navaid of top, landing stage, restricted area builds medium place.If its following sense of motion be point to bridge pier, restricted area, during to situations such as the visual plant facility of harbour, harbour critical area threaten, start warning device, give the alarm by broadcasting, remind the monitor staff and the marine navigator is given a warning, remind the deck officer to take to change course or speed change etc. dodged measure, avoids the generation of various accidents.

Installation system

System installs and should guarantee that thermal infrared imager 111 drives critical area and the waters that the visual field that forms can all cover needs monitoring by The Cloud Terrace 112, nothing is blocked, the installation site requires greater than more than 5 meters apart from the vertical distance of horizontal surface, and makes the imaging in whole covering critical areas and waters account for the over half of entire image zone.

Components of system as directed:

Referring to Fig. 1, Fig. 2, ship running intelligent all-weather actively safety forewarn system is made of target sampling unit 1, system's central processing unit 2, display unit 3, micro controller system 4, safe early warning unit 5;

Target sampling unit 1 is made up of the outdoor intelligent high-speed The Cloud Terrace 112 that has demoder, the thermal infrared imager 111 that has a video frequency collection card 113, thermal infrared imager 111 is fixed on the outdoor intelligent high-speed The Cloud Terrace 112, outdoor intelligent high-speed The Cloud Terrace 112 connects with system central processing unit 2 by demoder, and thermal infrared imager 111 connects with system central processing unit 2 by video frequency collection card 113;

Central processing system 2 comprises keyboard 26, Infrared video image digital signal processing unit 21, system decision-making processing unit 25;

Wherein,

Keyboard 26 is used for to system decision-making processing unit 25 input informations and control command;

Infrared video image digital signal processing unit 21, the data image signal that is used for receiving target sampling unit 1, according to data image signal, utilize processing and analytical equipment in the memory device, calculate position, azimuth, speed and the acceleration/accel of target, and result of calculation is delivered to system decision-making processing unit 25 usefulness by serial communication interface;

System decision-making processing unit 25 is used for and will comes from the information of digital signal processing unit, information in conjunction with the input of keyboard 26 information input units, carrying out the active safety early warning decision analyzes and calculates, obtain final dodge scheme after, carry out the active safety early warning simultaneously by display unit 3 and alarm unit 5;

Display unit 3, be used to show the native system installation place around target has been carried out the image of mark, and azimuth, speed and acceleration/accel between each target and the native system installation place show the scheme of dodging, the harmful grade of recommendation simultaneously;

Micro controller system 4 is used for the result of decision of receiving system decision-making treatment unit 25, the type of alarm of control alarm unit 5;

Alarm unit 5 is used to receive the control signal of micro controller system 4, uses the sound of different frequency, the information that the light mode is represented harmful grade and dangerous direction.

System is owing to adopt integrated design, realized real-time detection to distance, orientation, speed and acceleration/accel between system installation place and a plurality of motion ship target on every side, bridge, bridge pier, the reef target, the central processing unit intellectual analysis is handled, automatically carry out danger early warning and judge decision-making, and carry out auto alarm, and finally realize the purpose of boats and ships all-weather actively safety early warning by the mode that video, literal, sound, light, electricity etc. intuitively meet human features.

Infrared video image digital signal processing unit 21 is 2 floating-point signal processor TMS320C6713, and 2 TMS320C6713 are connected with the video frequency collection card 113 of thermal infrared imager 111; Described floating-point signal processor TMS320C6713 comprises video memory, programmable logic device, program store and controller, is provided with in the program store to handle and analytical equipment.It is little that TMS320C6713 has a volume, cost is low, characteristics such as stability is high, real-time is good, can realize the processing of 4 road video acquisitions simultaneously, TMS320C6713 has the alerting ability and the programmability of extremely strong processing capacity, height simultaneously, and is well positioned to meet the needs of target detect, tracking and recognizer.

Be provided with big view field image registration and splicing apparatus in the memory device in the Infrared video image digital signal processing unit 21, infrared sequence image is spliced into panoramic view automatically, and the panoramic view that will splice is automatically delivered to 25 storages of system decision-making processing unit.Big view field image registration and splicing apparatus utilize global motion estimating method, by the transformation parameter between estimated frames, have realized that the panoramic view of sequence image splices automatically.When kinematic parameter is estimated, adopted the block matching motion vector estimation of pyramid layering, effectively improve the running velocity of program, wherein add rejecting computing for abnormal mass, improve the precision of overall motion estimation greatly.In addition, under the situation that Illumination intensity between image differs.Adopt balanced equation of light computing, reached splicing effect preferably.This method can be accurately and fast the big view field image of generation video sequence, make the field range of video monitoring enlarge.The panoramic view of splicing is automatically delivered to 25 storages of system decision-making processing unit, can be used for post analysis evaluation and analysis on accident cause.

System decision-making processing unit 25 is a PC computing machine.Adopt common PC computing machine to have low price, powerful, interface convenience, can either satisfy the requirement that the system decision-making is calculated, be convenient to again be connected with each parts of system, simultaneously because the hard-disk capacity of computing machine is big, can fully store the panoramic view that image registration and splicing apparatus are spliced automatically in Infrared video image that target sampling unit 1 gathers and the Infrared video image digital signal processing unit 21.

Display unit 3 is 2 telltales, a plurality of target infrared video images of having marked around first telltale 31 and second telltale, 32, the first telltales 31 show; Relevant Word message: distance, azimuth, speed and acceleration/accel between ship running intelligent all-weather actively safety forewarn system installation place and a plurality of on every side boats and ships, bridge, bridge pier, the reef target; Second telltale 32 shows the text description of the scheme of dodging of recommending: comprise adopt speed change allow and/or come about allow, speed, orientation, harmful grade, dangerous direction, type of alarm.

Alarm unit 5 is, by controlling external loud speaker and/or alarm lamp with system decision-making processing unit 25 serial communication interface bonded assembly micro controller systems 4.

Ship running intelligent all-weather actively safety forewarn system can be fixedly mounted on bridge, harbour, harbour, landing stage, dangerous river section, restricted area, sluice gate or the ship.By system being fixedly mounted on ocean and harbour, interior korneforos, harbour, bridge, the landing stage, critical areas such as restricted area, utilize Infrared video image to handle the motion boats and ships by these critical areas are carried out detection and Identification, and it is carried out track and localization handle, judge the boats and ships sense of motion in future, the operation boats and ships are monitored in real time, when boats and ships enter prewarning area, through the computing machine judgment processing, if its following sense of motion is to point to bridge pier, the restricted area, to the harbour, during situations such as the visual plant facility of harbour critical area threatens, remind the monitor staff and the marine navigator is given a warning, avoid the generation of accident by broadcasting.

The method part:

Referring to Fig. 1, Fig. 2, Fig. 3, Fig. 4, the method for early warning of ship running intelligent all-weather actively safety forewarn system may further comprise the steps:

A, start-up system

Start ship running intelligent all-weather actively safety forewarn system, by keyboard 26 input informations and control command; Ship running intelligent all-weather actively safety forewarn system can be fixedly mounted on different positions, according to the size of actual conditions decisions prewarning area, and is provided with by keyboard 26, and by serial communication interface, delivers to system decision-making processing unit 25;

B, target sampling and digital signal processing

Sampling of Infrared video image target and digital signal processing: drive by The Cloud Terrace 112 with thermal infrared imager 111, to a/s time gap and angle step scan on every side, thereby make 111 pairs of aquatic environment on every side of thermal infrared imager make a video recording, is digital signal through video frequency collection card 113 with the image transitions of shooting with video-corder, and the programmable logic device of delivering in the Infrared video image digital signal processing unit 21 carries out sequential conversion and total line control, the video memory that control row signal and graphicinformation are delivered to respectively in the Infrared video image digital signal processing unit 21 is stored, video memory obtains giving after the information confirmation of programmable logic device in the Infrared video image digital signal processing unit 21, with processing in the program store in the Infrared video image digital signal processing unit 21 and analytical equipment, extracting image from video memory handles and analyzes, the result who handles and analyze is delivered to the controller of Infrared video image digital signal processing unit 21, after controller obtains information, give 113 1 acknowledgment signals of video frequency collection card, and final native system installation place and a plurality of targets on every side: boats and ships, bridge, bridge pier, azimuth between the reef, speed and acceleration information are delivered to system decision-making unit 25 by serial communication interface;

C, system decision-making unit 25 obtain the information from a, in conjunction with information from b, by carrying out that the active safety early warning decision is analyzed and calculate in system decision-making unit 25, obtain final dodge scheme after, carry out the active safety early warning simultaneously by telltale 3 and acousto-optic alarm terminal 5;

Wherein:

Telltale 3 content displayed are the image of a plurality of targets of having marked and relevant Word message around the native system installation place: azimuth, speed and acceleration/accel between native system installation place and a plurality of target boats and ships, bridge, bridge pier, the reef on every side; And the text description of the scheme of dodging of recommending: comprise adopt speed change allow and/or come about allow, speed, orientation;

Carry out the early warning of intelligent collision prevention active safety according to above-mentioned information, if it is dangerous, carrying out harmful grade judges, and by the sound of sound and light alarm terminal 5 use different frequencies, the information that the light mode is represented harmful grade and dangerous direction, by with system decision-making processing unit 25 serial communication interface bonded assembly micro controller systems 4, control external loud speaker or alarm lamp and report to the police.

In step b, also comprise image registration and splicing, infrared sequence image is spliced into panoramic view automatically, and the panoramic view that will splice is automatically delivered to 25 storages of system decision-making processing unit.Big view field image registration and splicing apparatus utilize global motion estimating method, by the transformation parameter between estimated frames, have realized that the panoramic view of sequence image splices automatically.When kinematic parameter is estimated, adopted the block matching motion vector estimation of pyramid layering, effectively improve the running velocity of program, wherein add rejecting computing for abnormal mass, improve the precision of overall motion estimation greatly.This method can be accurately and fast the big view field image of generation video sequence, make the field range of video monitoring enlarge.The panoramic view of splicing is automatically delivered to 25 storages of system decision-making processing unit, can be used for post analysis evaluation and analysis on accident cause.

Image processing and analysis are carried out according to following steps among the step b:

Carry out Infrared images pre-processing earlier, comprise image denoising, figure image intensifying and sharpening, image rectification; Infrared image is a set of being disturbed to form by real scene image, imaging noise and imaging.Suppose f (x, the y) infrared image that obtains of expression imaging system then include the target infrared scene image and can retouch speed and be:

f(x，y)＝f _T(x，y)+f _B(x，y)+n(x，y)+n ₁(x，y)

f _T(x y) is target gray value; f _B(x y) is background image, and (x y) is expressed as picture noise, n to n ₁(x y) is expressed as picture and disturbs.Background image f _B(x y) has long persistence length usually, and it has occupied scene image f (x, y) low-frequency information in the spatial frequency.Simultaneously, since the irregularity of scene and sensor internal heat distribution, background image f _B(x is a non-stationary process y), and local gray-value may have bigger variation in the image, in addition, and f _B(x y) also comprises high fdrequency component in the segment space frequency domain, and they mainly are distributed in the edge in each homogeneity district of Background image.

(x y) introduces in imaging process imaging noise n, is some microvariations that are superimposed upon on the infrared image random site; And n is disturbed in imaging ₁(x, y) then be since photosensitive unit response heterogeneity of infrared imaging or dummy argument cause it infrared image at random or form some wrong image data point on the fixed position.Therefore, n is disturbed in imaging ₁(x, y) in image, show as the pixel gray value much larger than or less than some isolated points of neighborhood intermediate value around it, the purpose of infrared image denoising is exactly to estimate the real scene image from image f.

Adopting two kinds of algorithms of different combinations to carry out sea horizon respectively extracts, at first original image being carried out picture quality estimates, determine whether image is carried out pretreatment according to evaluation result, carrying out sea horizon with two kinds of algorithms of different combinations simultaneously then extracts, wherein, the steps in sequence that the extraction algorithm combination that article one may sea horizon comprises is: the image iteration threshold is cut apart, Roberts gradient operator rim detection, refinement, utilize the Hough conversion to extract article one; Second may sea horizon the steps in sequence that comprises of extraction algorithm combination be: Roberts gradient operator rim detection and binaryzation, refinement, utilize the Hough conversion to extract the second sea horizon.After finishing two possible sea horizons extractions, get a final sea horizon that extracts of sea horizon conduct near the image lower end.

Image quality evaluation: adopt error of mean square (MSE) to come picture quality is estimated, be used for determining whether image is carried out pretreatment.

If MSE＞K then carries out the image pretreatment

If MSE≤K then does not carry out the image pretreatment

Get K=25.

First order image pretreatment: in the sky and ground region in infrared image more than the sea horizon, the existence of bridge, surface structures, continuous targets such as rock is arranged, the gray value of these targets generally all is in the high grade grey level of image, gradient between they and the surrounding environment is often greater than the gradient of sea horizon both sides, when especially these targets present the continuous linearity of nearly horizontal direction and distribute in image, cause present existed algorithms all can not correctly extract sea horizon.Adopting following method to eliminate the high target of these brightness values disturbs: the gray value of supposing pixel in the image is that (x, y), pixel adds up to M * N to f in the image, need the high pixel ratio of elimination brightness value be R, f _{M * N * R}(x, the y) gray value of the M * N * R pixel of expression when gray value is arranged from high to low, r represent pixel (x, eight lowest gray value in pixel y), behind the pretreatment pixel corresponding gray be g (x, y), then

if?f(x，y)≥f _M×N×R(x，y)then?g(x，y)＝r

if?f(x，y)＜f _M×N×R(x，y)then?g(x，y)＝f(x，y)

Owing to only reduce the brightness value of part high luminance values pixel, can the accurate extraction of sea horizon not impacted in this step.R gets 0.05.

Second stage image pretreatment: the water surface zone in infrared image below the sea horizon, the interference of strong ripples is to cause sea horizon can not extract another principal element.Strong ripples disturb at the shared pixel in water surface zone more, its grey value profile is near the average of entire image, and adopt following method to remove strong ripples and disturb: calculate image average fmean, the pixel corresponding gray is h (x behind the pretreatment of the second stage, y), then

if?f(x，y)＞fmean?then?h(x，y)＝f(x，y)-fmean

if?f(x，y)≤fmean?then?h(x，y)＝0

Behind the pretreatment of the second stage, most strong ripples disturb and are suppressed.Same this step can not impact the accurate extraction of sea horizon owing to only reduce the brightness value that the strong ripples of part disturb pixel.

The image iteration threshold is cut apart: entire image is regarded as by two zones and is constituted in the water transport environment, i.e. following water surface zone and sky and the ground region more than the sea horizon of sea horizon.When the MSE of original image≤K because the interference that is subjected to is little, directly image is carried out iteration threshold and cuts apart, can realize to two zones carry out actv., failure-free is cut apart; When the MSE of original image＞K, behind the image pretreatment, removed most of interference, at this moment again image is carried out iteration threshold and cuts apart, can realize to two zones carry out actv., failure-free is cut apart.

Roberts gradient operator rim detection: the Roberts gradient of asking for each picture element in the image respectively.

Binaryzation: gradient image is adopted edge threshold strategy binaryzation.In piece image, non-number of edge points is occupied certain proportion in total image pixel is counted out, and cooresponding factor of proportionality is expressed as Ratio.According to the corresponding histogram of image gradient value, begin progressively accumulated image point from low Grad grade, when the number that adds up reached image total pixel number purpose Ratio, cooresponding image gradient value was segmentation threshold.Ratio gets 0.95.

Refinement: have pixel in the bianry image and join together, influence the extraction precision and the real-time of Hough conversion, so will carry out refinement.The principle of refinement is for vertically being reduced to line segment a pixel.

Utilize the Hough conversion to extract sea horizon:

Utilize the Hough conversion to extract straight line, realized a kind ofly from the mapping of image space to parameter space, its basic thought is the duality of point one line.Polar equation by straight line: ρ=xcos θ+ysin θ carries out the Hough conversion, promptly uses the point on the cathetus of sine curve presentation video space.Herein parameter space is dispersed and change into an accumulator/accum battle array, with the every bit (x in the image, y) be mapped in the cooresponding a series of accumulator/accums of parameter space, make cooresponding accumulator value add 1, if comprise straight line in the image space, then in parameter space, there is a cooresponding accumulator/accum local maximum can occur.By detecting this local maximum, can determine that (ρ θ), thereby detects straight line with the cooresponding a pair of parameter of this straight line.

The reliability of sea horizon is the probabilistic problem of minimizing that a plurality of evidences of need are supported the same fact, with evidence theory the sea horizon that extracts is carried out confidence level for this reason and judges.

(1) whether sea horizon is positioned at possible sea horizon zone;

After native system was installed, must being in the sub regions of sea horizon in subregion, then extracted the confidence level M1=1 of sea horizon as the sea horizon that extracts, otherwise, extract the confidence level M1=0 of sea horizon.Pass through formula:

M1＝1，y1≤y≤y2

M1=0, y＜y1 or y＞y2

In the formula: y1, y2 are respectively the intermediate image vegetarian refreshments row-coordinate of the highest, minimum possibility sea horizon; Y is for extracting the intermediate image vegetarian refreshments row-coordinate of sea horizon.

(2) judge by the contrast ratio of sea horizon upper and lower subregion, obtain the confidence level of sea horizon;

Because the gray scale of sea horizon upper area generally will be higher than lower zone, when upper and lower region contrast contrast is big more, then the confidence level of sea horizon is high more.Calculate by formula:

${\mathrm{<figure-callout\; id="2"\; label="M"\; filenames="S2008100692312E00011.png,S2008100692312E00021.png"\; state="\{\{state\}\}">M</figure-callout>}}_2=\frac{[\underset{{\mathrm{<figure-callout\; id="1"\; label="h"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">h</figure-callout>}}_1<x<{\mathrm{<figure-callout\; id="2"\; label="h"\; filenames="S2008100692312E00011.png,S2008100692312E00021.png"\; state="\{\{state\}\}">h</figure-callout>}}_2}{\mathrm{\&Sigma;}}\underset{{\mathrm{<figure-callout\; id="1"\; label="w"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">w</figure-callout>}}_1<y<w_2}{\mathrm{\&Sigma;}}f(x,y)-\underset{{\mathrm{<figure-callout\; id="1"\; label="h"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">h</figure-callout>}}_1+\mathrm{\&Delta;h}<x<{\mathrm{<figure-callout\; id="2"\; label="h"\; filenames="S2008100692312E00011.png,S2008100692312E00021.png"\; state="\{\{state\}\}">h</figure-callout>}}_2+\mathrm{\&Delta;h}}{\mathrm{\&Sigma;}}\underset{{\mathrm{<figure-callout\; id="1"\; label="w"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">w</figure-callout>}}_1<y<w_2}{\mathrm{\&Sigma;}}f(x,y)]}{\underset{{\mathrm{<figure-callout\; id="1"\; label="h"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">h</figure-callout>}}_1<x<{\mathrm{<figure-callout\; id="2"\; label="h"\; filenames="S2008100692312E00011.png,S2008100692312E00021.png"\; state="\{\{state\}\}">h</figure-callout>}}_2}{\mathrm{\&Sigma;}}\underset{{\mathrm{<figure-callout\; id="1"\; label="w"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">w</figure-callout>}}_1<y<w_2}{\mathrm{\&Sigma;}}f(x,y)}$

The be associated confidence level M3 of evidence of the intermediate image vegetarian refreshments of the sea horizon that frame extracts (3).

${\mathrm{<figure-callout\; id="3"\; label="M"\; filenames="S2008100692312E00011.png"\; state="\{\{state\}\}">M</figure-callout>}}_3=\frac{1}{|y\left(t\right)-y(t-1)|},$ During y (t)=y (t-1), M ₃=1

Y (t), y (t-1) are respectively the intermediate image vegetarian refreshments row-coordinate of the sea horizon that extracts by present frame and former frame image, and when the difference of the two is big more, then the confidence level of sea horizon is low more.

(4) utilize the composition rule of D-S, obtain the comprehensive confidence level M of sea horizon, M=M is arranged ₁M ₂M ₃

Make that Mt is the thresholding of a suitable selection, when M＞Mt, think that the extraction of sea horizon of this two field picture is correct; Otherwise above-mentioned criterion can't be confirmed the correctness of the sea horizon of this frame.At this moment, preserve position, the regional up and down contrast ratio of sea horizon of former frame sea horizon.Proceeding the identification of the sea horizon of next frame handles.

Correct detect behind the sea horizon sea horizon up and down the interior image-region of certain limit as area-of-interest (ROI).

Interesting image regions ROI is carried out the evaluation of picture quality, when evaluation result is 1, adopt infrared target detection algorithm based on single-frame images; When evaluation result is 0, adopt infrared target detection algorithm based on image sequence;

Use the two-dimensional wavelet transformation in the wavelet analysis method to realize frequency selection and multiple dimensioned decomposition respectively, suppress background noise and strengthen target; Original image is carried out separating of low frequency part and HFS, respectively each low frequency and high fdrequency component are carried out multiresolution analysis then, extract target signature, carry out target detect;

In the fractal method according to the fractal characteristic of man-made targets such as boats and ships and bridge pier with the dimensional variation characteristics more violent than natural background, extract the Multi-scale Fractal image by the ROI subimage that carries out through the fuzzy filter method after the figure image intensifying; At last, utilize probabilistic relaxation that Multi-scale Fractal is carried out target detect; Fractal model be a kind of be suitable for describing have complicated and irregularly shaped mathematics model, its groundwork is to utilize natural scene and the man-made target difference on fractal dimension, the line style of logarithm value is estimated between the metric that common fractal dimension is each yardstick summer of calculating and the scale of measurement, think in each range scale that promptly fractal dimension is a constant value, this meets the feature of desirable fractal model.But for most of natural scenes, just in certain range scale, be similar to and have fractal characteristic, and real image is influenced by imaging noise, quantizing error etc., natural scene can not be described with the standard fractal dimension under many circumstances, and therefore the difference from the standard fractal dimension is distinguished natural background and man-made target often can not be obtained desirable effect.Consider that target detect is the fractal characteristic difference of utilizing between target and the background, do not need its fractal parameter of accurate Calculation, and background is different with the difference of dimensional variation with target in the image, utilizes the difference of both multiple dimensioned variations to carry out target detect.

Earlier the ROI subgraph is carried out medium filtering in the Mathematical Morphology Method, the pixel of getting brightness value maximum in the filtered image then image that serves as a mark; Original image is through top cap conversion, and the image after the row iteration threshold value of going forward side by side is cut apart carries out morphology reconstruct as mask images, realizes that infrared ship target detects; Morphology reconstruct: the thought of reconstruct is approached mask images by marking image is constantly expanded exactly, thereby recovers certain a part of or whole mask images of mask images, and this depends on choosing of marking image.The characteristics of reconstruct are by the choosing of marking image, and can extract own interesting areas in the mask images.

Define by following iterative process by f (marking image) reconstruct g (mask images):

H1 is initialized as marking image f;

Create 3x3 structural element B, each element all is 1 among the B, limits together with rule;

Repeat h _K+1=(h _k⊕ B) ∩ g is up to h _K+1=h _k

Attention: mark f must be the subclass of g, and the image after the reconstruct is the subclass of mask images g.

Use the evidential reasoning combination at last the target detect result from distinct methods is carried out the evidential reasoning combination, adopt Dempster evidence compositional rule that evidence is synthetic, obtain a high infrared ship target testing result of confidence level identification; Evidence theory can not need prior probability and conditional probability density, adopt belief function rather than probability as tolerance, " interval estimation " adopted in description to uncertain information, rather than the method for " point estimation ", differentiation do not know with uncertain aspect and reflect that accurately the evidence-gathering aspect demonstrates very big alerting ability.

Image processing and analysis are further comprising the steps of among the step b:

Using artificial neural networks technology, Fuzzy Inference realize the moving infrared ship target tracking of intelligent multimachine; Using artificial neural networks carries out the multimachine tracking of maneuvering target, has good adaptive, self-organized learning and association and fault-tolerant ability, by study and training, stronger judgement and discernibility is arranged, and can independently find the way of dealing with problems; Use Fuzzy Inference to a plurality of targets of being followed the tracks of assessment that impends, judgement target type, target velocity, thus infer their threat level; Artificial neural net (ANN) is compared with traditional adaptive system at random, by study and training, stronger judgement and discernibility is arranged, and can independently find the way of dealing with problems.And its fault-tolerant ability is very strong, under data and environment uncertain condition, especially under the situation of much noise and interference, still can carry out target detect, parameter estimation, target's feature-extraction and identification, system modelling etc.Fuzzy reasoning is to have imitated the reasoning process of human thinking to real things; the multiple target tracking process " may be related " or " may be not related " between the target occur through regular meeting, and target " may belong to a class target " or analogues such as " may not belong to a class target ".The utilization fuzzy reasoning is judged target type, target velocity etc. to a plurality of targets of being followed the tracks of assessment that impends, thereby infers their threat level, can accurately propose the collision avoidance scheme for grasping the ship personnel, provides a favorable security.The traditional method of these utilizations is difficult to satisfy the demand on the engineering, and the utilization fuzzy reasoning can simplify a problem, to satisfy the needs of real-time in the target tracking.

Image processing and analysis are further comprising the steps of among the step b:

Employing increases self adaptation and learning ability in the classical mode recognizer, utilize the image context, the knowledge base that incorporates artificial intelligence technology carries out the identification of infrared ship target, carry out the extraction of target signature in cut zone, difference extracting position feature, shape facility, size characteristic, radiation feature and the feature of extracting based on wavelet analysis, and, carry out the identification of infrared ship target the input end of these features as the RBF neural network.Design-calculated RBF neural network of the present invention has the three-layer network structure, and input layer comprises 8 input parameters, and hidden layer has 12 nodes, the type of output layer input target.

Among the step c, according to the information of importing among the step a, distance, azimuth, speed and acceleration/accel between the native system installation place that obtains among the integrating step b and a plurality of on every side target boats and ships, bridge, bridge pier, the reef; Adopt the fuzzy expert system method,, comprehensively become the principal component of separate quadrature, finally determine the ship collision prevention degree of risk, carry out the active early warning decision based on this each early warning decision key element and influence degree thereof; If native system is mounted on the ship, according to this ship model that obtains among the step a, size, this ship handling parameter, load-carrying, this ship position, the real-time ship's speed of this ship, distance, azimuth, speed and acceleration/accel between the native system installation place that obtains among the integrating step b and a plurality of on every side target boats and ships, bridge, bridge pier, the reef; Adopt the fuzzy expert system method,, comprehensively become the principal component of separate quadrature, finally determine the ship collision prevention degree of risk, carry out the active early warning decision based on this each early warning decision key element and influence degree thereof.

The active safety forewarn system is by complication system that people-ship-the environment three elements are formed.Thereby the factor that influences people, boats and ships, environment three elements produces significant effects to the active safety early warning decision.On the basis of the collision prevention information of utilizing marine navigation equipment to obtain, intend adopting the fuzzy expert system method to determine the ship collision prevention degree of risk, carry out the active early warning decision based on this.

Boats and ships under steam, it is x that each sensor obtains the early warning decision key element ₁, x ₂, x ₃X _n, each key element is different to the influence degree of early warning decision, and also interacts between the key element.

If i key element is w to the influence degree of early warning decision _i(1≤i≤n), n sample establishing extraction is X:x ₁ ^t, x ₂ ^t..., x _n ^t((1≤t≤n)). n characteristic index comprehensively becomes independently principal component y of n mutually orthogonal now ₁, y ₂..., y _n, write as matrix form and be

Y＝C·X (2)

Wherein

$C=\left[\begin{array}{ccc}{\mathrm{<figure-callout\; id="11"\; label="C"\; filenames="S2008100692312E00011.png"\; state="\{\{state\}\}">C</figure-callout>}}_{11}& ...& C_{1n}\\ ...& ...& ...\\ {\mathrm{<figure-callout\; id="1"\; label="C\; n"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">C</figure-callout>}}_n& ...& C_{\mathrm{nn}}\end{array}\right],$ $Y=\left[\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="y"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">y</figure-callout>}}_1\\ \&CenterDot;\&CenterDot;\&CenterDot;\\ y_n\end{array}\right]$

Fuzzy expert system can be handled probabilistic data and proposition, promptly can be in [0,1] middle value, and its adopts fuzzy technology means such as fuzzy set, fuzzy number and fuzzy relation to represent and handle the uncertainty and the inexactness of knowledge.Because in the ship running process, uncertain factor is very many, so the active safety early warning decision adopts the practicality of fuzzy expert system energy enhanced system.

The active safety forewarn system is made up of three main portions such as information collection, data handing, decision-makings.

Embodiment 2: system is installed on the boats and ships

Ship running intelligent all-weather actively safety forewarn system is fixedly mounted on the place ahead, ship-handling chamber, by boats and ships a plurality of target boats and ships on every side to native system is installed, bridge, bridge pier, reef carries out detection and Identification, and it is carried out track and localization handle, monitor in real time, on the basis that obtains collision prevention information, and through a plurality of target boats and ships around the automatic judgement of computing machine, bridge, bridge pier, whether reef constitutes a threat to the boats and ships that native system is installed, adopt the fuzzy expert system method to determine the ship collision prevention degree of risk, carry out the active early warning decision based on this.

Certainly,, also can be fixedly mounted on system the front portion, top layer roofing deck of boats and ships according to the difference of field condition, or roll-on-roll-off ship the place such as choose above the arm.When around a plurality of target boats and ships, bridge, bridge pier, reef during to situations such as the boats and ships that native system is installed threaten, the comprehensive mode of using video, sound, light, electricity etc. to meet human features is carried out auto alarm, the marine navigator is given a warning, avoid the generation of accident.Can greatly improve the perception of behaviour's ship personnel for the navigation surrounding environment, improve the active safety navigation supportability of boats and ships under the low visibility situation, auxiliary behaviour's ship personnel make a strategic decision, reduce the error of behaviour's ship, improve initiatively collision prevention success ratio of boats and ships greatly, support personnel's the life and the safety of property reduce or avoid the generation of gross pollution waters and natural environment accident, guarantee to navigate by water transportation safety.

Installation system

System installs and should guarantee that thermal infrared imager 111 drives critical area and the waters that the visual field that forms can all cover needs monitoring by The Cloud Terrace 112, nothing is blocked, the installation site requires greater than more than 5 meters apart from the vertical distance of horizontal surface, and makes the imaging in whole covering critical areas and waters account for the over half of entire image zone.

Components of system as directed:

Referring to Fig. 1, Fig. 2, ship running intelligent all-weather actively safety forewarn system is made of target sampling unit 1, system's central processing unit 2, display unit 3, micro controller system 4, safe early warning unit 5;

Target sampling unit 1 is made up of the outdoor intelligent high-speed The Cloud Terrace 112 that has demoder, the thermal infrared imager 111 that has a video frequency collection card 113, thermal infrared imager 111 is fixed on the outdoor intelligent high-speed The Cloud Terrace 112, outdoor intelligent high-speed The Cloud Terrace 112 connects with system central processing unit 2 by demoder, and thermal infrared imager 111 connects with system central processing unit 2 by video frequency collection card 113;

Central processing system 2 comprises keyboard 26, Infrared video image digital signal processing unit 21, system decision-making processing unit 25;

Wherein,

Keyboard 26 is used for to system decision-making processing unit 25 input informations and control command;

Infrared video image digital signal processing unit 21, the data image signal that is used for receiving target sampling unit 1, according to data image signal, utilize processing and analytical equipment in the memory device, calculate position, azimuth, speed and the acceleration/accel of target, and result of calculation is delivered to system decision-making processing unit 25 usefulness by serial communication interface;

System decision-making processing unit 25 is used for and will comes from the information of digital signal processing unit, information in conjunction with the input of keyboard 26 information input units, carrying out the active safety early warning decision analyzes and calculates, obtain final dodge scheme after, carry out the active safety early warning simultaneously by display unit 3 and alarm unit 5;

Display unit 3, be used to show the native system installation place around target has been carried out the image of mark, and azimuth, speed and acceleration/accel between each target and the native system installation place show the scheme of dodging, the harmful grade of recommendation simultaneously;

Micro controller system 4 is used for the result of decision of receiving system decision-making treatment unit 25, the type of alarm of control alarm unit 5;

Alarm unit 5 is used to receive the control signal of micro controller system 4, uses the sound of different frequency, the information that the light mode is represented harmful grade and dangerous direction.

System is owing to adopt integrated design, realized real-time detection to distance, orientation, speed and acceleration/accel between system installation place and a plurality of motion ship target on every side, bridge, bridge pier, the reef target, the central processing unit intellectual analysis is handled, automatically carry out danger early warning and judge decision-making, and carry out auto alarm, and finally realize the purpose of boats and ships all-weather actively safety early warning by the mode that video, literal, sound, light, electricity etc. intuitively meet human features.

Infrared video image digital signal processing unit 21 is 2 floating-point signal processor TMS320C6713, and 2 TMS320C6713 are connected with the video frequency collection card 113 of thermal infrared imager 111; Described floating-point signal processor TMS320C6713 comprises video memory, programmable logic device, program store and controller, is provided with in the program store to handle and analytical equipment.It is little that TMS320C6713 has a volume, cost is low, characteristics such as stability is high, real-time is good, can realize the processing of 4 road video acquisitions simultaneously, TMS320C6713 has the alerting ability and the programmability of extremely strong processing capacity, height simultaneously, and is well positioned to meet the needs of target detect, tracking and recognizer.

Be provided with big view field image registration and splicing apparatus in the memory device in the Infrared video image digital signal processing unit 21, infrared sequence image is spliced into panoramic view automatically, and the panoramic view that will splice is automatically delivered to the storage of system decision-making processing unit.Big view field image registration and splicing apparatus utilize global motion estimating method, by the transformation parameter between estimated frames, have realized that the panoramic view of sequence image splices automatically.When kinematic parameter is estimated, adopted the block matching motion vector estimation of pyramid layering, effectively improve the running velocity of program, wherein add rejecting computing for abnormal mass, improve the precision of overall motion estimation greatly.In addition, under the situation that Illumination intensity between image differs.Adopt balanced equation of light computing, reached splicing effect preferably.This method can be accurately and fast the big view field image of generation video sequence, make the field range of video monitoring enlarge.The panoramic view of splicing is automatically delivered to 25 storages of system decision-making processing unit, can be used for post analysis evaluation and analysis on accident cause.

System decision-making processing unit 25 is a PC computing machine.Adopt common PC computing machine to have low price, powerful, interface convenience, can either satisfy the requirement that the system decision-making is calculated, be convenient to again be connected with each parts of system, simultaneously because the hard-disk capacity of computing machine is big, can fully store the panoramic view that the Infrared video image of target sampling unit collection and image registration and splicing apparatus splice automatically.Display unit 3 is 2 telltales, the image of a plurality of targets of having marked around first telltale 31 and second telltale, 32, the first telltales 31 show; Relevant Word message: distance, azimuth, speed and acceleration/accel between ship running intelligent all-weather actively safety forewarn system installation place and a plurality of on every side boats and ships, bridge, bridge pier, the reef target; Second telltale 32 shows the text description of the scheme of dodging of recommending: comprise adopt speed change allow and/or come about allow, speed, orientation, harmful grade, dangerous direction, type of alarm.Adopt the display mode of two telltales, utilize to have obtained real-time panoramic view image, and can meet multiple goals such as boats and ships (a plurality of target), bridge (bridge pier), water surface reef and carry out actv. detection, recognition and tracking other of the boats and ships front region of travelling.By water surface zone and detected a plurality of target of splicing later panoramic view image are carried out colored painted Rendering Flags target, and simultaneously the panoramic view image that splices later panoramic view image and the colored painted Rendering Flags target of process is simultaneously displayed on first telltale 31, image demonstration through mark can guarantee that the user intuitively uses native system, utilize text description to satisfy user's different demands so that the user is convenient to the performance of system is estimated.

Alarm unit 5 is, by controlling external loud speaker and/or alarm lamp with system decision-making processing unit 25 serial communication interface bonded assembly micro controller systems 4.The comprehensive mode of using video, sound, light, electricity etc. to meet human features is carried out auto alarm, improves the perception of behaviour's ship personnel for navigation environment, and the collision avoidance decision-making is assisted, thereby improves the ship collision prevention success ratio.

Target sampling unit 1 also is equipped with radar 121, be provided with the visible images sensor 122 of band video frequency collection card 123 before radar 121 screens, visible images sensor 122 connects with visible light video image digital signal processing unit 22 in the system central processing unit 2 by video frequency collection card 123, and described visible light video image digital signal processing unit 22 is 2 floating-point signal processor TMS320C6713; Described floating-point signal processor TMS320C6713 comprises video memory, programmable logic device, program store and controller, is provided with in the program store to handle and analytical equipment; Also be provided with information fusion unit 23 at described central processing system 2, information fusion unit 23, be used for the various information that come from Infrared video image digital signal processing unit 21 and visible light video image digital signal processing unit 22 are respectively carried out information fusion, obtain distance, azimuth, speed and the acceleration/accel of ultimate aim, simultaneously the target in the image is made a mark, and result of calculation is delivered to system decision-making processing unit 25 by serial communication interface; Described information fusion unit 23 is a high-speed real-time digital signal processor ADSP21060.Do not destroy former marine radar navigationsystem in this way, do not influence the use of radar 121, do not change electric wiring, only install visible light sensor additional on former radar 121, multiobject information brings great convenience in the radar 121 in order to obtain.The inner integrated dual-port static memory device of 4Mbit of ADSP21060 has special-purpose peripheral I/O bus, and the radical function with digital information processing system is integrated on a slice chip very effectively, constitutes monolithic system easily, dwindles the volume of circuit card.High speed instruction buffer memory device (CACHE) in the sheet makes call instruction carry out with pipeline system, guarantee that every instruction finishes in the monocycle (25ns), floating-point peak value arithmetic speed is up to 120MFLOPS (million floating point operations per second), and normal floating point operation speed is 80MFLOPS.The controller of peripheral I/O bus provides six cover high-speed link mouths and two cover synchronous serial interfaces, a large amount of ADSP21060 can be constituted a loose coupling parallel processing system (PPS) with link port.In addition, treater comprises that also 3 interrupt pin, 4 sign pins, MSO-MS3 sheets select pin, make treater and other Peripheral Interfaces simple.In a word, the address space that ADSP21060 is huge, powerful addressing mode, the very long instruction word of 48bit and floating point operation ability thereof can satisfy the demand of fusion device software algorithm real-time fully.

Display unit 3 also shows this ship image of a plurality of targets of having marked on every side; Relevant Word message: distance, azimuth, speed and acceleration/accel between this ship model, size, this ship handling parameter, load-carrying, this ship position, the real-time ship's speed of this ship, this ship and a plurality of on every side boats and ships, bridge, bridge pier, the reef target.

The method part:

Referring to Fig. 1, Fig. 2, Fig. 3, Fig. 4, the method for early warning of ship running intelligent all-weather actively safety forewarn system may further comprise the steps:

A, start-up system

Start ship running intelligent all-weather actively safety forewarn system, by keyboard 26 input informations and control command; By boats and ships model, dimension of ship, boats and ships handling parameter and the weight information of keyboard 26 these ships of input, this ship position among this ship boat-carrying GPS24 and real-time ship's speed information,, deliver to system decision-making processing unit 25 simultaneously by serial communication interface;

B, target sampling and sampling of digital signal processing Infrared video image target and digital signal processing: drive by The Cloud Terrace 112 with thermal infrared imager 111, to a/s time gap and angle step scan on every side, thereby make 111 pairs of aquatic environment on every side of thermal infrared imager make a video recording, is digital signal through video frequency collection card 113 with the image transitions of shooting with video-corder, and the programmable logic device of delivering in the Infrared video image digital signal processing unit 21 carries out sequential conversion and total line control, the video memory that control row signal and graphicinformation are delivered to respectively in the Infrared video image digital signal processing unit 21 is stored, video memory obtains giving after the information confirmation of programmable logic device in the Infrared video image digital signal processing unit, with processing in the program store in the Infrared video image digital signal processing unit 21 and analytical equipment, extracting image from video memory handles and analyzes, the result who handles and analyze is delivered to the controller of Infrared video image digital signal processing unit 21, after controller obtains information, give 113 1 acknowledgment signals of video frequency collection card, and final native system installation place and a plurality of targets on every side: boats and ships, bridge, bridge pier, azimuth between the reef, speed and acceleration information are delivered to system decision-making unit 25 by serial communication interface;

C, system decision-making unit 25 obtain the information from a, in conjunction with information from b, by carrying out that the active safety early warning decision is analyzed and calculate in system decision-making unit 25, obtain final dodge scheme after, carry out the active safety early warning simultaneously by telltale 3 and acousto-optic alarm terminal 5;

Wherein:

Telltale 3 content displayed are the image of a plurality of targets of having marked and relevant Word message around the native system installation place: distance, azimuth, speed and acceleration/accel between native system installation place and a plurality of target boats and ships, bridge, bridge pier, the reef on every side; If native system is mounted on the ship, also should show this ship model, size, this ship handling parameter, load-carrying, this ship position, the real-time ship's speed of this ship and the text description of the scheme of dodging of recommending: comprise adopt speed change allow and/or come about allow, speed, orientation;

Carry out the early warning of intelligent collision prevention active safety according to above-mentioned information, if it is dangerous, carrying out harmful grade judges, and by the sound of sound and light alarm terminal 5 use different frequencies, the information that the light mode is represented harmful grade and dangerous direction, by with system decision-making processing unit 25 serial communication interface bonded assembly micro controller systems 4, control external loud speaker or alarm lamp and report to the police.

In step b, also comprise image registration and splicing, sequence image is spliced into panoramic view automatically, and the panoramic view that will splice is automatically delivered to 25 storages of system decision-making processing unit.Big view field image registration and splicing apparatus utilize global motion estimating method, by the transformation parameter between estimated frames, have realized that the panoramic view of sequence image splices automatically.When kinematic parameter is estimated, adopted the block matching motion vector estimation of pyramid layering, effectively improve the running velocity of program, wherein add rejecting computing for abnormal mass, improve the precision of overall motion estimation greatly.This method can be accurately and fast the big view field image of generation video sequence, make the field range of video monitoring enlarge.The panoramic view of splicing is automatically delivered to 25 storages of system decision-making processing unit, can be used for post analysis evaluation and analysis on accident cause.

Image processing and analysis are carried out according to following steps among the step b:

Carry out Infrared images pre-processing earlier, comprise image denoising, figure image intensifying and sharpening, image rectification, movement background correction; Infrared image is a set of being disturbed to form by real scene image, imaging noise and imaging.Suppose f (x, the y) infrared image that obtains of expression imaging system then include the target infrared scene image and can retouch speed and be:

f(x，y)＝f _T(x，y)+f _B(x，y)+n(x，y)+n ₁(x，y)

f _T(x y) is target gray value; f _B(x y) is background image, and (x y) is expressed as picture noise, n to n ₁(x y) is expressed as picture and disturbs.Background image f _B(x y) has long persistence length usually, and it has occupied scene image f (x, y) low-frequency information in the spatial frequency.Simultaneously, since the irregularity of scene and sensor internal heat distribution, background image f _B(x is a non-stationary process y), and local gray-value may have bigger variation in the image, in addition, and f _B(x y) also comprises high fdrequency component in the segment space frequency domain, and they mainly are distributed in the edge in each homogeneity district of Background image.

(x y) introduces in imaging process imaging noise n, is some microvariations that are superimposed upon on the infrared image random site; And n is disturbed in imaging ₁(x, y) then be since photosensitive unit response heterogeneity of infrared imaging or dummy argument cause it infrared image at random or form some wrong image data point on the fixed position.Therefore, n is disturbed in imaging ₁(x, y) in image, show as the pixel gray value much larger than or less than some isolated points of neighborhood intermediate value around it, the purpose of infrared image denoising is exactly to estimate the real scene image from image f.

Movement background is proofreaied and correct because imageing sensor is installed on the motion platform sometimes, even or be installed on the static platform, also may be because some interference causes the sensor shake, the shake of sensor will cause background shake.By the global motion parameter estimation techniques.It at first estimates the kinematic parameter of background, utilizes the kinematic parameter that estimates to do background correction then, and multiple image is proofreaied and correct under the same coordinate system.

Directly, extract bridge, bridge pier and reef feature then, carry out actv. and follow the tracks of and discern image region segmentation;

Adopting two kinds of algorithms of different combinations to carry out sea horizon respectively extracts, wherein, the steps in sequence that the extraction algorithm combination that article one may sea horizon comprises is: the image iteration threshold is cut apart, Roberts gradient operator rim detection, refinement, utilize the Hough conversion to extract article one sea horizon; Second may sea horizon the steps in sequence that comprises of extraction algorithm combination be: Roberts gradient operator rim detection and binaryzation, refinement, utilize the Hough conversion to extract the second sea horizon; Get in two possible sea horizons a sea horizon near the image lower end as the final sea horizon that extracts, and sea horizon carried out confidence level judge, then sea horizon up and down the image-region in the certain limit as region of interest ROI;

Based on the position relation of sea horizon and ship target, the ship target image generally is in the sea horizon zone.This is determined by the middle distance planar imaging.Target can not break away from sea horizon fully and be in a day dummy section, also can not break away from sea horizon fully and is in other zones such as land, valley.Therefore the correct sea horizon that detects, then sea horizon up and down the image-region in the certain limit as area-of-interest (ROI), the detection of follow-up infrared ship target, tracking and identification, just dwindled the image processing scope greatly, and greatly avoided the interference of some high-radiation areas on aerial cloud, ripples, land or the valley, thereby make that the calculated amount of various algorithms is significantly reduced, guaranteed the requirement of algorithm real-time.

The sea horizon leaching process that the present invention adopts is: at first original image is carried out picture quality and estimate, determine whether image is carried out pretreatment according to evaluation result, carrying out sea horizon with two kinds of algorithms of different combinations simultaneously then extracts, wherein, the steps in sequence that the extraction algorithm combination that article one may sea horizon comprises is: the image iteration threshold is cut apart, Roberts gradient operator rim detection, refinement, utilize the Hough conversion to extract article one; Second may sea horizon the steps in sequence that comprises of extraction algorithm combination be: Roberts gradient operator rim detection and binaryzation, refinement, utilize the Hough conversion to extract the second sea horizon.After finishing two possible sea horizons extractions, get a final sea horizon that extracts of sea horizon conduct near the image lower end.

Image quality evaluation: adopt error of mean square (MSE) to come picture quality is estimated, be used for determining whether image is carried out pretreatment.

If MSE＞K then carries out the image pretreatment

If MSE≤K then does not carry out the image pretreatment

Get K=25.

First order image pretreatment: in the sky and ground region in infrared image more than the sea horizon, the existence of bridge, surface structures, continuous targets such as rock is arranged, the gray value of these targets generally all is in the high grade grey level of image, gradient between they and the surrounding environment is often greater than the gradient of sea horizon both sides, when especially these targets present the continuous linearity of nearly horizontal direction and distribute in image, cause present existed algorithms all can not correctly extract sea horizon.Adopting following method to eliminate the high target of these brightness values disturbs: the gray value of supposing pixel in the image is that (x, y), pixel adds up to M * N to f in the image, need the high pixel ratio of elimination brightness value be R, f _{M * N * R}(x, the y) gray value of the M * N * R pixel of expression when gray value is arranged from high to low, r represent pixel (x, eight lowest gray value in pixel y), behind the pretreatment pixel corresponding gray be g (x, y), then

if?f(x，y)≥f _M×N×R(x，y)then?g(x，y)＝r

if?f(x，y)＜f _M×N×R(x，y)then?g(x，y)＝f(x，y)

Owing to only reduce the brightness value of part high luminance values pixel, can the accurate extraction of sea horizon not impacted in this step.R gets 0.05.

Second stage image pretreatment: the water surface zone in infrared image below the sea horizon, the interference of strong ripples is to cause sea horizon can not extract another principal element.Strong ripples disturb at the shared pixel in water surface zone more, its grey value profile is near the average of entire image, and adopt following method to remove strong ripples and disturb: calculate image average fmean, the pixel corresponding gray is h (x behind the pretreatment of the second stage, y), then

if?f(x，y)＞fmean?then?h(x，y)＝f(x，y)-fmean

if?f(x，y)≤fmean?then?h(x，y)＝0

Behind the pretreatment of the second stage, most strong ripples disturb and are suppressed.Same this step can not impact the accurate extraction of sea horizon owing to only reduce the brightness value that the strong ripples of part disturb pixel.

The image iteration threshold is cut apart: entire image is regarded as by two zones and is constituted in the water transport environment, i.e. following water surface zone and sky and the ground region more than the sea horizon of sea horizon.When the MSE of original image≤K because the interference that is subjected to is little, directly image is carried out iteration threshold and cuts apart, can realize to two zones carry out actv., failure-free is cut apart; When the MSE of original image＞K, behind the image pretreatment, removed most of interference, at this moment again image is carried out iteration threshold and cuts apart, can realize to two zones carry out actv., failure-free is cut apart.

Roberts gradient operator rim detection: the Roberts gradient of asking for each picture element in the image respectively.

Binaryzation: gradient image is adopted edge threshold strategy binaryzation.In piece image, non-number of edge points is occupied certain proportion in total image pixel is counted out, and cooresponding factor of proportionality is expressed as Ratio.According to the corresponding histogram of image gradient value, begin progressively accumulated image point from low Grad grade, when the number that adds up reached image total pixel number purpose Ratio, cooresponding image gradient value was segmentation threshold.Ratio gets 0.95.

Refinement: have pixel in the bianry image and join together, influence the extraction precision and the real-time of Hough conversion, so will carry out refinement.The principle of refinement is for vertically being reduced to line segment a pixel.

Utilize the Hough conversion to extract sea horizon:

Utilize the Hough conversion to extract straight line, realized a kind ofly from the mapping of image space to parameter space, its basic thought is the duality of point one line.Polar equation by straight line: ρ=xcos θ+ysin θ carries out the Hough conversion, promptly uses the point on the cathetus of sine curve presentation video space.Herein parameter space is dispersed and change into an accumulator/accum battle array, with the every bit (x in the image, y) be mapped in the cooresponding a series of accumulator/accums of parameter space, make cooresponding accumulator value add 1, if comprise straight line in the image space, then in parameter space, there is a cooresponding accumulator/accum local maximum can occur.By detecting this local maximum, can determine that (ρ θ), thereby detects straight line with the cooresponding a pair of parameter of this straight line.

The reliability of sea horizon is the probabilistic problem of minimizing that a plurality of evidences of need are supported the same fact, with evidence theory the sea horizon that extracts is carried out confidence level for this reason and judges.

(1) whether sea horizon is positioned at possible sea horizon zone;

When native system after installing on the boats and ships, must being in the sub regions of sea horizon in subregion, then extracted the confidence level M1=1 of sea horizon as the sea horizon that extracts, otherwise, extract the confidence level M1=0 of sea horizon.Pass through formula:

M1＝1，y1≤y≤y2

M1=0, y＜y1 or y＞y2

In the formula: y1, y2 are respectively the intermediate image vegetarian refreshments row-coordinate of the highest, minimum possibility sea horizon; Y is for extracting the intermediate image vegetarian refreshments row-coordinate of sea horizon.

(2) judge by the contrast ratio of sea horizon upper and lower subregion, obtain the confidence level of sea horizon;

Because the gray scale of sea horizon upper area generally will be higher than lower zone, when upper and lower region contrast contrast is big more, then the confidence level of sea horizon is high more.Calculate by formula:

${\mathrm{<figure-callout\; id="2"\; label="M"\; filenames="S2008100692312E00011.png,S2008100692312E00021.png"\; state="\{\{state\}\}">M</figure-callout>}}_2=\frac{[\underset{{\mathrm{<figure-callout\; id="1"\; label="h"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">h</figure-callout>}}_1<x<{\mathrm{<figure-callout\; id="2"\; label="h"\; filenames="S2008100692312E00011.png,S2008100692312E00021.png"\; state="\{\{state\}\}">h</figure-callout>}}_2}{\mathrm{\&Sigma;}}\underset{{\mathrm{<figure-callout\; id="1"\; label="w"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">w</figure-callout>}}_1<y<w_2}{\mathrm{\&Sigma;}}f(x,y)-\underset{{\mathrm{<figure-callout\; id="1"\; label="h"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">h</figure-callout>}}_1+\mathrm{\&Delta;h}<x<{\mathrm{<figure-callout\; id="2"\; label="h"\; filenames="S2008100692312E00011.png,S2008100692312E00021.png"\; state="\{\{state\}\}">h</figure-callout>}}_2+\mathrm{\&Delta;h}}{\mathrm{\&Sigma;}}\underset{{\mathrm{<figure-callout\; id="1"\; label="w"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">w</figure-callout>}}_1<y<w_2}{\mathrm{\&Sigma;}}f(x,y)]}{\underset{{\mathrm{<figure-callout\; id="1"\; label="h"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">h</figure-callout>}}_1<x<{\mathrm{<figure-callout\; id="2"\; label="h"\; filenames="S2008100692312E00011.png,S2008100692312E00021.png"\; state="\{\{state\}\}">h</figure-callout>}}_2}{\mathrm{\&Sigma;}}\underset{{\mathrm{<figure-callout\; id="1"\; label="w"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">w</figure-callout>}}_1<y<w_2}{\mathrm{\&Sigma;}}f(x,y)}$

The be associated confidence level M3 of evidence of the intermediate image vegetarian refreshments of the sea horizon that frame extracts (3).

${\mathrm{<figure-callout\; id="3"\; label="M"\; filenames="S2008100692312E00011.png"\; state="\{\{state\}\}">M</figure-callout>}}_3=\frac{1}{|y\left(t\right)-y(t-1)|},$ During y (t)=y (t-1), M ₃=1

Y (t), y (t-1) are respectively the intermediate image vegetarian refreshments row-coordinate of the sea horizon that extracts by present frame and former frame image, and when the difference of the two is big more, then the confidence level of sea horizon is low more.

(4) utilize the composition rule of D-S, obtain the comprehensive confidence level M of sea horizon, M=M is arranged ₁M ₂M ₃

Make that Mt is the thresholding of a suitable selection, when M＞Mt, think that the extraction of sea horizon of this two field picture is correct; Otherwise above-mentioned criterion can't be confirmed the correctness of the sea horizon of this frame.At this moment, preserve position, the regional up and down contrast ratio of sea horizon of former frame sea horizon.Proceeding the identification of the sea horizon of next frame handles.

Correct detect behind the sea horizon sea horizon up and down the interior image-region of certain limit as area-of-interest (ROI).

Interesting image regions ROI is carried out the evaluation of picture quality, when evaluation result is 1, adopt infrared target detection algorithm based on single-frame images; When evaluation result is 0, adopt infrared target detection algorithm based on image sequence;

Use the two-dimensional wavelet transformation in the wavelet analysis method to realize frequency selection and multiple dimensioned decomposition respectively, suppress background noise and strengthen target; Original image is carried out separating of low frequency part and HFS, respectively each low frequency and high fdrequency component are carried out multiresolution analysis then, extract target signature, carry out target detect;

In the fractal method according to the fractal characteristic of man-made targets such as boats and ships and bridge pier with the dimensional variation characteristics more violent than natural background, extract the Multi-scale Fractal image by the ROI subimage that carries out through the fuzzy filter method after the figure image intensifying; At last, utilize probabilistic relaxation that Multi-scale Fractal is carried out target detect; Fractal model be a kind of be suitable for describing have complicated and irregularly shaped mathematics model, its groundwork is to utilize natural scene and the man-made target difference on fractal dimension, the line style of logarithm value is estimated between the metric that common fractal dimension is each yardstick summer of calculating and the scale of measurement, think in each range scale that promptly fractal dimension is a constant value, this meets the feature of desirable fractal model.But for most of natural scenes, just in certain range scale, be similar to and have fractal characteristic, and real image is influenced by imaging noise, quantizing error etc., natural scene can not be described with the standard fractal dimension under many circumstances, and therefore the difference from the standard fractal dimension is distinguished natural background and man-made target often can not be obtained desirable effect.Consider that target detect is the fractal characteristic difference of utilizing between target and the background, do not need its fractal parameter of accurate Calculation, and background is different with the difference of dimensional variation with target in the image, utilizes the difference of both multiple dimensioned variations to carry out target detect.

Earlier the ROI subgraph is carried out medium filtering in the Mathematical Morphology Method, the pixel of getting brightness value maximum in the filtered image then image that serves as a mark; Original image is through top cap conversion, and the image after the row iteration threshold value of going forward side by side is cut apart carries out morphology reconstruct as mask images, realizes that infrared ship target detects; Morphology reconstruct: the thought of reconstruct is approached mask images by marking image is constantly expanded exactly, thereby recovers certain a part of or whole mask images of mask images, and this depends on choosing of marking image.The characteristics of reconstruct are by the choosing of marking image, and can extract own interesting areas in the mask images.

Define by following iterative process by f (marking image) reconstruct g (mask images):

H1 is initialized as marking image f;

Create 3x3 structural element B, each element all is 1 among the B, limits together with rule;

Repeat h _K+1=(h _k⊕ B) ∩ g is up to h _K+1=h _k

Attention: mark f must be the subclass of g, and the image after the reconstruct is the subclass of mask images g.

Use the evidential reasoning combination at last the target detect result from distinct methods is carried out the evidential reasoning combination, adopt Dempster evidence compositional rule that evidence is synthetic, obtain a high infrared ship target testing result of confidence level identification; Evidence theory can not need prior probability and conditional probability density, adopt belief function rather than probability as tolerance, " interval estimation " adopted in description to uncertain information, rather than the method for " point estimation ", differentiation do not know with uncertain aspect and reflect that accurately the evidence-gathering aspect demonstrates very big alerting ability.

Image processing and analysis are further comprising the steps of among the step b:

Using artificial neural networks technology, Fuzzy Inference realize the moving infrared ship target tracking of intelligent multimachine; Using artificial neural networks carries out the multimachine tracking of maneuvering target, has good adaptive, self-organized learning and association and fault-tolerant ability, by study and training, stronger judgement and discernibility is arranged, and can independently find the way of dealing with problems; Use Fuzzy Inference to a plurality of targets of being followed the tracks of assessment that impends, judgement target type, target velocity, thus infer their threat level; Artificial neural net (ANN) is compared with traditional adaptive system at random, by study and training, stronger judgement and discernibility is arranged, and can independently find the way of dealing with problems.And its fault-tolerant ability is very strong, under data and environment uncertain condition, especially under the situation of much noise and interference, still can carry out target detect, parameter estimation, target's feature-extraction and identification, system modelling etc.Fuzzy reasoning is to have imitated the reasoning process of human thinking to real things; the multiple target tracking process " may be related " or " may be not related " between the target occur through regular meeting, and target " may belong to a class target " or analogues such as " may not belong to a class target ".The utilization fuzzy reasoning is judged target type, target velocity etc. to a plurality of targets of being followed the tracks of assessment that impends, thereby infers their threat level, can accurately propose the collision avoidance scheme for grasping the ship personnel, provides a favorable security.The traditional method of these utilizations is difficult to satisfy the demand on the engineering, and the utilization fuzzy reasoning can simplify a problem, to satisfy the needs of real-time in the target tracking.

Image processing and analysis are further comprising the steps of among the step b:

Employing increases self adaptation and learning ability in the classical mode recognizer, utilize the image context, the knowledge base that incorporates artificial intelligence technology carries out the identification of infrared ship target, carry out the extraction of target signature in cut zone, difference extracting position feature, shape facility, size characteristic, radiation feature and the feature of extracting based on wavelet analysis, and, carry out the identification of infrared ship target the input end of these features as the RBF neural network.Design-calculated RBF neural network of the present invention has the three-layer network structure, and input layer comprises 8 input parameters, and hidden layer has 12 nodes, the type of output layer input target.

Further comprising the steps of among the step b:

B1, sampling of visible light video image target and digital signal processing: with the visible images sensor 122 that is installed in radar 121 screen fronts, radar 121 screens are made a video recording, obtain radar 121 screen pictures, is digital signal through video frequency collection card 123 with the image transitions of shooting with video-corder, and the programmable logic device of delivering in the visible light video image digital signal processing unit 22 carries out sequential conversion and total line control, the video memory that control row signal and graphicinformation are delivered to respectively in the visible light video image digital signal processing unit 22 is stored, video memory obtains giving after the information confirmation of programmable logic device in the visible light video image digital signal processing unit 22, with processing in the program store in the visible light video image digital signal processing unit 22 and analytical equipment, extracting image from video memory handles and analyzes, the result who handles and analyze is delivered to the controller of visible light video image digital signal processing unit 22, after controller obtains information, give 123 1 acknowledgment signals of video frequency collection card, and final native system installation place and a plurality of targets on every side: boats and ships, bridge, bridge pier, distance between the reef, the azimuth, speed and acceleration information, the information fusion unit of delivering in the central processing system 2 by serial communication interface 23 carries out the breath fusion;

B2, information fusion: information fusion unit 23 obtains coming from the information of Infrared video image digital signal processing unit 21 and visible light video image digital signal processing unit 22, give Infrared video image digital signal processing unit 21 and 22 1 confirmations of visible light video image digital signal processing unit among the corresponding step b respectively, information fusion unit 23 will Infrared video image digital signal processing unit 21 obtains from step b native system installation place and a plurality of target boats and ships on every side, bridge, bridge pier, azimuth between the reef, speed and acceleration/accel, and the native system installation place that obtains of visible light video image digital signal processing unit 22 and a plurality of target boats and ships on every side, bridge, distance between the reef, the azimuth, speed and acceleration/accel, handle through information fusion, obtain final fusion results information, be the native system installation place and a plurality of target boats and ships on every side, bridge, bridge pier, distance between the reef, the azimuth, the accurate information of speed and acceleration/accel, simultaneously the target in the image is made a mark, deliver to system decision-making unit 25 by serial communication interface.

According to the information of importing among the step a, distance, azimuth, speed and acceleration/accel between the native system installation place that obtains among the integrating step b and a plurality of on every side target boats and ships, bridge, bridge pier, the reef; Adopt the fuzzy expert system method,, comprehensively become the principal component of separate quadrature, finally determine the ship collision prevention degree of risk, carry out the active early warning decision based on this each early warning decision key element and influence degree thereof; If native system is mounted on the ship, according to this ship model that obtains among the step a, size, this ship handling parameter, load-carrying, this ship position, the real-time ship's speed of this ship, distance, azimuth, speed and acceleration/accel between the native system installation place that obtains among the integrating step b and a plurality of on every side target boats and ships, bridge, bridge pier, the reef; Adopt the fuzzy expert system method,, comprehensively become the principal component of separate quadrature, finally determine the ship collision prevention degree of risk, carry out the active early warning decision based on this each early warning decision key element and influence degree thereof.

The active safety forewarn system is by complication system that people-ship-the environment three elements are formed.Thereby the factor that influences people, boats and ships, environment three elements produces significant effects to the active safety early warning decision.On the basis of the collision prevention information of utilizing marine navigation equipment to obtain, intend adopting the fuzzy expert system method to determine the ship collision prevention degree of risk, carry out the active early warning decision based on this.

Boats and ships under steam, it is x that each sensor obtains the early warning decision key element ₁, x ₂, x ₃X _n, each key element is different to the influence degree of early warning decision, and also interacts between the key element.

If i key element is w to the influence degree of early warning decision _i(1≤i≤n), n sample establishing extraction is X:x ₁ ^t, x ₂ ^t..., x _n ^t((1≤t≤n)). n characteristic index comprehensively becomes independently principal component y of n mutually orthogonal now ₁, y ₂..., y _n, write as matrix form and be

Y＝C·X (2)

Wherein

$C=\left[\begin{array}{ccc}{\mathrm{<figure-callout\; id="11"\; label="C"\; filenames="S2008100692312E00011.png"\; state="\{\{state\}\}">C</figure-callout>}}_{11}& ...& C_{1n}\\ ...& ...& ...\\ {\mathrm{<figure-callout\; id="1"\; label="C\; n"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">C</figure-callout>}}_n& ...& C_{\mathrm{nn}}\end{array}\right],$ $Y=\left[\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="y"\; filenames="S2008100692312E00021.png"\; state="\{\{state\}\}">y</figure-callout>}}_1\\ \&CenterDot;\&CenterDot;\&CenterDot;\\ y_n\end{array}\right]$

Fuzzy expert system can be handled probabilistic data and proposition, promptly can be in [0,1] middle value, and its adopts fuzzy technology means such as fuzzy set, fuzzy number and fuzzy relation to represent and handle the uncertainty and the inexactness of knowledge.Because in the ship running process, uncertain factor is very many, so the active safety early warning decision adopts the practicality of fuzzy expert system energy enhanced system.

The active safety forewarn system is made up of three main portions such as information collection, data handing, decision-makings.The a plurality of target boats and ships in this ship running intelligent all-weather actively safety forewarn system that obtains according to the Infrared video image digital signal processing unit 21 among the step b among the step b2 and the place ahead, bridge, bridge pier, azimuth between the reef, speed and acceleration/accel, the a plurality of target boats and ships of this ship running intelligent all-weather actively safety forewarn system and the place ahead that obtain with visible light video image digital signal processing unit 22 from step b1, bridge, distance between the reef, the azimuth, speed and acceleration/accel, utilize thermal infrared imager 111 high-precision measurement of angle and radar 121 high-precision range observations, utilize message complementary sense, by information fusion technology, provide accurate estimation to the target location; The employing centralized processing method that merges characteristic layer realizes that radar 121 and thermal infrared imager 111 merge, at first extract the target barycenter in the infrared image, with the least-squares estimation method the redundant angle technology of thermal infrared imager 111 is compressed then, measure the pseudo-measurement of angle of aiming to produce in time with radar 121, carry out fusion treatment with the measurement of azimuth of radar 121 respectively again, merge estimation to obtain synchrodata, at last merging the dbjective state that the data that obtain are used to upgrade filter based on radar 121 and thermal infrared imager 111, the fusion process of decision-making level adopts distributed approach, at first set up flight path separately, and then carry out radar 121 and the related of thermal infrared imager 111 flight paths and merge about target by radar 121 and thermal infrared imager 111.

Radar 121 is as active sensor, because can round-the-clock measurement and provide target complete location information, thereby bringing into play important effect aspect target acquisition and the tracking.But because radar 121 will be to aerial radiation large power, electrically magnetic wave in when work, thereby subjects to electro countermeasure, and angle-measurement accuracy is lower.

Thermal infrared imager 111 is not to any energy of aerial radiation, it is surveyed by the heat energy of receiving target radiation and locatees, have stronger antijamming capability, simultaneously thermal infrared imager 111 also has advantages such as the high and target recognition ability of angle accuracy is strong, but can not find range from.

Utilize radar 121 high-precision range observations and thermal infrared imager 111 high-precision measurement of angle, utilize message complementary sense,, can provide accurate estimation, improve tracking and identification target to the target location by information fusion technology.

(1) sensor measurement model

What thermal infrared imager 111 was measured is the azimuth and the angular altitude at object brightness center, and the brightness center of hypothetical target overlaps with barycenter here, and then its measurement model is:

${\mathrm{\&theta;}}_I\left(k\right)=\mathrm{\&theta;}\left(k\right)+{\mathrm{\&upsi;}}_{{\mathrm{\&theta;}}_I}\left(k\right),$

${\mathrm{\&phi;}}_I\left(k\right)=\mathrm{\&phi;}\left(k\right)+{\mathrm{\&upsi;}}_{{\mathrm{\&phi;}}_I}\left(k\right).$

θ in the formula _I(k), φ _I(k) be infrared observed reading, θ (k), φ (k) are actual angle,

Be the measurement of angle noise, their average is zero, variance is a Gaussian white noise.The dbjective state vector is elected position, speed and the acceleration/accel in the inertial system as, promptly

$X\left(k\right)={\left[x\stackrel{\&CenterDot;}{\left(k\right)\stackrel{\&CenterDot;}{x}}\left(k\right)\stackrel{\&CenterDot;\&CenterDot;}{x}\left(k\right)y\left(k\right)\stackrel{\&CenterDot;}{y}\left(k\right)\stackrel{\&CenterDot;\&CenterDot;}{y}\left(k\right)z\left(k\right)\stackrel{\&CenterDot;}{z}\left(k\right)\stackrel{\&CenterDot;\&CenterDot;}{z}\left(k\right)\right]}^T,$ Then have

$\left[\begin{array}{c}{\mathrm{\&theta;}}_I\left(k\right)\\ {\mathrm{\&phi;}}_I\left(k\right)\end{array}\right]=\left[\begin{array}{c}\arctan [z\left(k\right)/x\left(k\right)]\\ \arctan [y\left(k\right)/\sqrt{x^2\left(k\right)+z^2\left(k\right)}]\end{array}\right]+\left[\begin{array}{c}{\mathrm{\&upsi;}}_{{\mathrm{\&theta;}}_I}\left(k\right)\\ {\mathrm{\&upsi;}}_{{\mathrm{\&phi;}}_I}\left(k\right)\end{array}\right].$

Radar 121 is the distance and the azimuth of measurement target directly, and its measurement model is

$r_R\left(k\right)=r\left(k\right)+{\mathrm{\&upsi;}}_{r_R}\left(k\right),$

${\mathrm{\&theta;}}_R\left(k\right)=\mathrm{\&theta;}\left(k\right)+{\mathrm{\&upsi;}}_{{\mathrm{\&theta;}}_R}\left(k\right),$

${\mathrm{\&phi;}}_R\left(k\right)=\mathrm{\&phi;}\left(k\right)+{\mathrm{\&upsi;}}_{{\mathrm{\&phi;}}_R}\left(k\right).$

R in the formula _R(k), θ _R(k), φ _R(k) be the observed reading of radar 121, r (k), θ (k), φ (k) are actual value,

Be to measure noise, their average is zero, variance is a Gaussian white noise.The dbjective state vector is chosen the same, then has

$\left[\begin{array}{c}{r}_R\left(k\right)\\ {\mathrm{\&theta;}}_R\left(k\right)\\ {\mathrm{\&phi;}}_R\left(k\right)\end{array}\right]=\left[\begin{array}{c}\sqrt{x^2\left(k\right)+y^2\left(k\right)+z^2\left(k\right)}\\ \arctan [z\left(k\right)/x\left(k\right)]\\ \arctan [y\left(k\right)/\sqrt{x^2\left(k\right)+z^2\left(k\right)}]\end{array}\right]+\left[\begin{array}{c}{\mathrm{\&upsi;}}_{r_R}\left(k\right)\\ {\mathrm{\&upsi;}}_{{\mathrm{\&theta;}}_R}\left(k\right)\\ {\mathrm{\&upsi;}}_{{\mathrm{\&phi;}}_R}\left(k\right)\end{array}\right].$

(2) structure mode and the algorithm of radar, infrared image signal fusion and decision system

What characteristic layer merged act as: utilize the characteristic information of IR imaging target identification and tracking to help the target recognition and the tracking of radar, thereby improve the target detection probability of system and reduce false-alarm probability; What decision-making level merged act as: when target relative distance is far away, guide the servo control unit tracking target of thermal infrared imager 111 according to the tracking decision information of radar 121 modules, target is dropped in the visual angle of thermal infrared imager 111, thermal infrared imager 111 can be discerned and tracking target voluntarily by imaging analysis during near target with box lunch, thereby can remedy the near deficiency of thermal infrared imager 111 ranging coverages, performance thermal infrared imager 111 is followed the tracks of the high advantage of precision of decision information near target the time.When losing tracking target ability or tracking target ability because of reason one sensor assemblies such as interference, can correct the target tracking ability of this sensor assembly that is disturbed according to the tracking decision information of another sensor assembly, thereby improve the anti-interference resistance of system, improved the reliability of whole target recognition track channel in addition, in case because of software or the wherein a certain sensor of hardware fault have lost target recognition and traceability, the fusion center decision controller still can and be followed the tracks of the correct tracking target of decision signal according to the target recognition of another sensor.

Because the data collection rate of thermal infrared imager 111 is apparently higher than radar 121, therefore, characteristic layer merges the centralized processing method that adopts, realize that the basic thought that radar 121 and thermal infrared imager 111 merge is: at first extract the target barycenter in the infrared image, with the least-squares estimation method the redundant angle technology of thermal infrared imager 111 is compressed then, measure the pseudo-measurement of angle of aiming to produce in time with radar 121, carry out fusion treatment with the measurement of azimuth of radar 121 respectively again, merge estimation to obtain synchrodata, at last merging the dbjective state that the data that obtain are used to upgrade filter based on radar 121 and thermal infrared imager 111.

Merge radar and infrared data fusion for decision-making level, adopt distributed approach,, and then carry out radar and the related of infrared flight path and fusion promptly at first by radar and the infrared flight path of setting up separately about target.

Claims (17)

1. ship running intelligent all-weather actively safety forewarn system is characterized in that: be made of target sampling unit (1), system's central processing unit (2), display unit (3), micro controller system (4), safe early warning unit (5);

Described target sampling unit (1) is made up of the outdoor intelligent high-speed The Cloud Terrace (112) that has demoder, the thermal infrared imager (111) that has video frequency collection card (113), thermal infrared imager (111) is fixed on the outdoor intelligent high-speed The Cloud Terrace (112), outdoor intelligent high-speed The Cloud Terrace (112) connects with system's central processing unit (2) by demoder, and thermal infrared imager (111) connects with system's central processing unit (2) by video frequency collection card (113);

Described central processing system (2) comprises keyboard (26), Infrared video image digital signal processing unit (21), system decision-making processing unit (25);

Wherein,

Keyboard (26) is used for to system decision-making processing unit (25) input information and control command;

Infrared video image digital signal processing unit (21), the data image signal that is used for receiving target sampling unit (1), according to data image signal, utilize processing and analytical equipment in the memory device, calculate position, azimuth, speed and the acceleration/accel of target, and by serial communication interface result of calculation is delivered to system decision-making processing unit (25) and use;

System decision-making processing unit (25) is used for and will comes from the information of digital signal processing unit, information in conjunction with the input of keyboard (26) information input unit, carrying out the active safety early warning decision analyzes and calculates, obtain final dodge scheme after, carry out the active safety early warning simultaneously by display unit (3) and safe early warning unit (5);

Display unit (3), be used to show the native system installation place around target has been carried out the image of mark, and azimuth, speed and acceleration/accel between each target and the native system installation place show the scheme of dodging, the harmful grade of recommendation simultaneously;

Micro controller system (4) is used for the result of decision of receiving system decision-making treatment unit (25), the type of alarm of control safe early warning unit (5);

Safe early warning unit (5) is used to receive the control signal of micro controller system (4), uses the sound of different frequency, the information that the light mode is represented harmful grade and dangerous direction.

2. ship running intelligent all-weather actively safety forewarn system according to claim 1, it is characterized in that: described Infrared video image digital signal processing unit (21) is 2 floating-point signal processor TMS320C6713, and 2 TMS320C6713 are connected with the video frequency collection card (113) of thermal infrared imager (111); Described floating-point signal processor TMS320C6713 comprises video memory, programmable logic device, program store and controller, is provided with in the program store to handle and analytical equipment.

3. ship running intelligent all-weather actively safety forewarn system according to claim 1, it is characterized in that: be provided with big view field image registration and splicing apparatus in the memory device in described Infrared video image digital signal processing unit (21), infrared sequence image is spliced into panoramic view automatically, and the panoramic view that will splice is automatically delivered to system decision-making processing unit (25) storage.

4. ship running intelligent all-weather actively safety forewarn system according to claim 1 is characterized in that: described system decision-making processing unit (25) is a PC computing machine.

5. ship running intelligent all-weather actively safety forewarn system according to claim 1, it is characterized in that: described display unit (3) is 2 telltales, first telltale (31) and second telltale (32), a plurality of target infrared video images of having marked and relevant Word message around first telltale (31) shows: distance, azimuth, speed and acceleration/accel between ship running intelligent all-weather actively safety forewarn system installation place and a plurality of on every side boats and ships, bridge, bridge pier, the reef target; The text description of the scheme of dodging that second telltale (32) show to be recommended: comprise adopt speed change allow and/or come about allow, speed, orientation, harmful grade, dangerous direction, type of alarm.

6. ship running intelligent all-weather actively safety forewarn system according to claim 1 is characterized in that: described safe early warning unit (5) is by controlling external loud speaker and/or alarm lamp with system decision-making processing unit (25) serial communication interface bonded assembly micro controller system (4).

7. ship running intelligent all-weather actively safety forewarn system according to claim 1, it is characterized in that: described target sampling unit (1) also is equipped with radar (121), be provided with the visible images sensor (122) of band video frequency collection card (123) before radar (121) screen, visible images sensor (122) connects with visible light video image digital signal processing unit (22) in system's central processing unit (2) by video frequency collection card (123), and described visible light video image digital signal processing unit (22) is 2 floating-point signal processor TMS320C6713; Described floating-point signal processor TMS320C6713 comprises video memory, programmable logic device, program store and controller, is provided with in the program store to handle and analytical equipment; Also be provided with information fusion unit (23) at described central processing system (2), information fusion unit (23), be used for the various information that come from Infrared video image digital signal processing unit (21) and visible light video image digital signal processing unit (22) are respectively carried out information fusion, obtain distance, azimuth, speed and the acceleration/accel of ultimate aim, simultaneously the target in the image is made a mark, and result of calculation is delivered to system decision-making processing unit (25) by serial communication interface; Described information fusion unit (23) is a high-speed real-time digital signal processor ADSP21060.

8. ship running intelligent all-weather actively safety forewarn system according to claim 1 is characterized in that: described ship running intelligent all-weather actively safety forewarn system is provided in a side of on bridge, harbour, harbour, landing stage, dangerous river section, restricted area, sluice gate or the ship; When this ship running intelligent all-weather actively safety forewarn system was established aboard ship, display unit (3) also showed the image of a plurality of targets of having marked around this ship and relevant Word message: distance, azimuth, speed and acceleration/accel between this ship model, size, this ship handling parameter, load-carrying, this ship position, the real-time ship's speed of this ship, this ship and a plurality of on every side boats and ships, bridge, bridge pier, the reef target.

9. realize the method for early warning of the described ship running intelligent all-weather actively safety of claim 1 forewarn system, it is characterized in that may further comprise the steps:

A, start-up system

Start ship running intelligent all-weather actively safety forewarn system, by keyboard (26) input information and control command; If ship running intelligent all-weather actively safety forewarn system is mounted on the ship, then import boats and ships model, dimension of ship, boats and ships handling parameter and the weight information of this ship by keyboard (26), simultaneously this ship position among this ship boat-carrying GPS (24) and real-time ship's speed information, by serial communication interface, deliver to system decision-making processing unit (25);

B, target sampling and digital signal processing

Sampling of Infrared video image target and digital signal processing: drive by The Cloud Terrace (112) with thermal infrared imager (111), to a/s time gap and angle step scan on every side, thereby make thermal infrared imager (111) make a video recording to aquatic environment on every side, is digital signal through video frequency collection card (113) with the image transitions of shooting with video-corder, and the programmable logic device of delivering in the Infrared video image digital signal processing unit (21) carries out sequential conversion and total line control, the video memory that control row signal and graphicinformation are delivered to respectively in the Infrared video image digital signal processing unit (21) is stored, video memory obtains giving after the information confirmation of programmable logic device in the Infrared video image digital signal processing unit (21), with processing and the analytical equipment in the program store in the Infrared video image digital signal processing unit (21), extracting image from video memory handles and analyzes, the result who handles and analyze is delivered to the controller of Infrared video image digital signal processing unit (21), after controller obtains information, give (113) acknowledgment signals of video frequency collection card, and final native system installation place and a plurality of targets on every side: boats and ships, bridge, bridge pier, azimuth between the reef, speed and acceleration information are delivered to system decision-making processing unit (25) by serial communication interface;

C, system decision-making processing unit (25) obtain the information from a, in conjunction with information from b, carrying out the active safety early warning decision by system decision-making processing unit (25) analyzes and calculates, obtain final dodge scheme after, carry out the active safety early warning simultaneously by telltale (3) and acousto-optic safe early warning unit (5);

Wherein:

Telltale (3) content displayed is the image of a plurality of targets of having marked and relevant Word message around the native system installation place: azimuth, speed and acceleration/accel between native system installation place and a plurality of target boats and ships, bridge, bridge pier, the reef on every side; If native system is mounted on the ship, also should show this ship model, size, this ship handling parameter, load-carrying, this ship position, the real-time ship's speed of this ship and the text description of the scheme of dodging of recommending: comprise adopt speed change allow and/or come about allow, speed, orientation;

Carry out the early warning of intelligent collision prevention active safety according to above-mentioned information, if it is dangerous, carrying out harmful grade judges, and by the sound of acousto-optic safe early warning unit (5) use different frequency, the information that the light mode is represented harmful grade and dangerous direction, by with system decision-making processing unit (25) serial communication interface bonded assembly micro controller system (4), control external loud speaker or alarm lamp and report to the police.

10. method for early warning according to claim 9 is characterized in that: also comprise image registration and splicing infrared sequence image being spliced into panoramic view automatically in step b, and the panoramic view that will splice is automatically delivered to system decision-making processing unit (25) storage.

11. method for early warning according to claim 9 is characterized in that image processing is carried out according to following steps with analyzing among the step b:

Carry out Infrared images pre-processing earlier, comprise image denoising, figure image intensifying and sharpening, image rectification, movement background correction;

Directly, extract bridge, bridge pier and reef feature then, carry out actv. and follow the tracks of and discern image region segmentation;

Adopting two kinds of algorithms of different combinations to carry out sea horizon respectively extracts, wherein, the steps in sequence that the extraction algorithm combination that article one may sea horizon comprises is: the image iteration threshold is cut apart, Roberts gradient operator rim detection, refinement, utilize the Hough conversion to extract article one sea horizon; Second may sea horizon the steps in sequence that comprises of extraction algorithm combination be: Roberts gradient operator rim detection and binaryzation, refinement, utilize the Hough conversion to extract the second sea horizon; Get in two possible sea horizons a sea horizon near the image lower end as the final sea horizon that extracts, and sea horizon carried out confidence level judge, then sea horizon up and down the image-region in the certain limit as region of interest ROI;

Interesting image regions ROI is carried out the evaluation of picture quality, when evaluation result is 1, adopt infrared target detection algorithm based on single-frame images; When evaluation result is 0, adopt infrared target detection algorithm based on image sequence;

Use the two-dimensional wavelet transformation in the wavelet analysis method to realize frequency selection and multiple dimensioned decomposition respectively, suppress background noise and strengthen target; Original image is carried out separating of low frequency part and HFS, respectively each low frequency and high fdrequency component are carried out multiresolution analysis then, extract target signature, carry out target detect; In the fractal method according to the fractal characteristic of man-made target with the dimensional variation characteristics more violent than natural background, extract the Multi-scale Fractal image by the ROI subimage that carries out through the fuzzy filter method after the figure image intensifying; At last,

Utilize probabilistic relaxation that Multi-scale Fractal is carried out target detect; Earlier the ROI subgraph is carried out medium filtering in the Mathematical Morphology Method, the pixel of getting brightness value maximum in the filtered image then image that serves as a mark; Original image is through top cap conversion, and the image after the row iteration threshold value of going forward side by side is cut apart carries out morphology reconstruct as mask images, realizes that infrared ship target detects; Use the evidential reasoning combination at last the target detect result from distinct methods is carried out the evidential reasoning combination, adopt Dempster evidence compositional rule that evidence is synthetic, obtain a high infrared ship target testing result of confidence level identification.

12. method for early warning according to claim 9 is characterized in that: image processing and analysis are further comprising the steps of among the step b:

Using artificial neural networks technology, Fuzzy Inference realize the moving infrared ship target tracking of intelligent multimachine; Using artificial neural networks carries out the multimachine tracking of maneuvering target, has good adaptive, self-organized learning and association and fault-tolerant ability, by study and training, stronger judgement and discernibility is arranged, and can independently find the way of dealing with problems; Use Fuzzy Inference to a plurality of targets of being followed the tracks of assessment that impends, judgement target type, target velocity, thus infer their threat level.

13. method for early warning according to claim 9 is characterized in that: image processing and analysis are further comprising the steps of among the step b:

Employing increases self adaptation and learning ability in the classical mode recognizer, utilize the image context, the knowledge base that incorporates artificial intelligence technology carries out the identification of infrared ship target, carry out the extraction of target signature in cut zone, difference extracting position feature, shape facility, size characteristic, radiation feature and the feature of extracting based on wavelet analysis, and, carry out the identification of infrared ship target the input end of these features as the RBF neural network.

14. method for early warning according to claim 9 is characterized in that: further comprising the steps of among the step b:

B1, sampling of visible light video image target and digital signal processing: with the visible images sensor (122) that is installed in radar (121) screen front, radar (121) screen is made a video recording, obtain radar (121) screen picture, is digital signal through video frequency collection card (123) with the image transitions of shooting with video-corder, and the programmable logic device of delivering in the visible light video image digital signal processing unit (22) carries out sequential conversion and total line control, the video memory that control row signal and graphicinformation are delivered to respectively in the visible light video image digital signal processing unit (22) is stored, video memory obtains giving after the information confirmation of programmable logic device in the visible light video image digital signal processing unit (22), with processing and the analytical equipment in the program store in the visible light video image digital signal processing unit (22), extracting image from video memory handles and analyzes, the result who handles and analyze is delivered to the controller of visible light video image digital signal processing unit (22), after controller obtains information, give (123) acknowledgment signals of video frequency collection card, and final native system installation place and a plurality of targets on every side: boats and ships, bridge, bridge pier, distance between the reef, the azimuth, speed and acceleration information, the information fusion unit of delivering in the central processing system (2) by serial communication interface (23) carries out information fusion;

B2, information fusion: information fusion unit (23) obtain coming from the information of Infrared video image digital signal processing unit (21) and visible light video image digital signal processing unit (22), give Infrared video image digital signal processing unit (21) and (22) confirmations of visible light video image digital signal processing unit among the corresponding step b respectively, information fusion unit (23) will Infrared video image digital signal processing unit (21) obtains from step b native system installation place and a plurality of target boats and ships on every side, bridge, bridge pier, azimuth between the reef, speed and acceleration/accel, and the native system installation place that obtains of visible light video image digital signal processing unit (22) and a plurality of target boats and ships on every side, bridge, distance between the reef, the azimuth, speed and acceleration/accel, handle through information fusion, obtain final fusion results information, be the native system installation place and a plurality of target boats and ships on every side, bridge, bridge pier, distance between the reef, the azimuth, the accurate information of speed and acceleration/accel, simultaneously the target in the image is made a mark, deliver to system decision-making unit (25) by serial communication interface.

15. method for early warning according to claim 9, it is characterized in that: according to the information of importing among the step a, distance, azimuth, speed and acceleration/accel between the native system installation place that obtains among the integrating step b and a plurality of on every side target boats and ships, bridge, bridge pier, the reef; Adopt the fuzzy expert system method,, comprehensively become the principal component of separate quadrature, finally determine the ship collision prevention degree of risk, carry out the active early warning decision based on this each early warning decision key element and influence degree thereof.

16. method for early warning according to claim 9, it is characterized in that: native system is mounted on the ship, according to this ship model that obtains among the step a, size, this ship handling parameter, load-carrying, this ship position, the real-time ship's speed of this ship, distance, azimuth, speed and acceleration/accel between the native system installation place that obtains among the integrating step b and a plurality of on every side target boats and ships, bridge, bridge pier, the reef; Adopt the fuzzy expert system method,, comprehensively become the principal component of separate quadrature, finally determine the ship collision prevention degree of risk, carry out the active early warning decision based on this each early warning decision key element and influence degree thereof.

17. method for early warning according to claim 14, it is characterized in that: a plurality of target boats and ships of this ship running intelligent all-weather actively safety forewarn system that obtains according to the Infrared video image digital signal processing unit (21) among the step b among the step b2 and the place ahead, bridge, bridge pier, azimuth between the reef, speed and acceleration/accel, the a plurality of target boats and ships of this ship running intelligent all-weather actively safety forewarn system and the place ahead that obtain with visible light video image digital signal processing unit (22) from step b1, bridge, distance between the reef, the azimuth, speed and acceleration/accel, utilize high-precision measurement of angle of thermal infrared imager (111) and the high-precision range observation of radar (121), utilize message complementary sense, by information fusion technology, provide accurate estimation to the target location; The employing centralized processing method that merges characteristic layer realizes that radar (121) and thermal infrared imager (111) merge, at first extract the target barycenter in the infrared image, use the least-squares estimation method that the redundant angle technology of thermal infrared imager (111) is compressed then, with the pseudo-measurement of angle that produces in time and radar (121) is measured aligning, carry out fusion treatment with the measurement of azimuth of radar (121) respectively again, merge estimation to obtain synchrodata, at last merging the dbjective state that the data that obtain are used to upgrade filter based on radar (121) and thermal infrared imager (111), the fusion process of decision-making level adopts distributed approach, at first set up flight path separately, and then carry out radar (121) and the related of thermal infrared imager (111) flight path and merge about target by radar (121) and thermal infrared imager (111).

Images