CN101447075A - Wide-angle lens-based FPGA & DSP embedded multi-valued targets threshold categorization tracking device - Google Patents
Wide-angle lens-based FPGA & DSP embedded multi-valued targets threshold categorization tracking device Download PDFInfo
- Publication number
- CN101447075A CN101447075A CNA2008101547392A CN200810154739A CN101447075A CN 101447075 A CN101447075 A CN 101447075A CN A2008101547392 A CNA2008101547392 A CN A2008101547392A CN 200810154739 A CN200810154739 A CN 200810154739A CN 101447075 A CN101447075 A CN 101447075A
- Authority
- CN
- China
- Prior art keywords
- tracking
- target
- dsp
- image
- valued
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention provides a wide-angle lens-based FPGA & DSP embedded multi-valued targets threshold categorization tracking device and relates to an embedded system for identifying and tracking multi-targets in a video stream and a related algorithm. Image collection is completed by a wide-angle lens and a color area array CMOS chip; digital image pretreatment, such as digital filtering, image enhancement and the like, is carried out by the FPGA; the algorithms such as multi-valued targets threshold categorization identification, marking registration and the like are realized in a main processor with the DSP as a core; an improved image-tracking program which is based on multi-targets cross operation of a probabilistic forecasting model generates a tracking gate in real time; a target tester in the tracking gate controls and tracks a process and outputs a target value. The wide-angle lens-based multi-valued targets threshold categorization tracking device supported by an embedded hardware platform has wide application prospect in the aspects of dynamic photography, security monitoring, maneuvering target detecting, multi-targets tracking, automatic navigation of vehicles, etc. The device especially has the advantage in constructing an airborne target tracking system with small structure volume and low power consumption.
Description
[technical field]:
The invention belongs to embedded machine vision uses and the target following technical field.
[background technology]:
Common image tracker mostly obtains target information by conventional camera lens.Because the field angle of conventional camera lens is less, can only obtain on-the-spot limited local message, cause the blind area of vision system bigger, be unfavorable for the real-time follow-up of target.The advantage that the overall view visual system that the present invention adopts big wide-angle lens to make up has big visual field whole audience scape to shoot with video-corder, on the one hand, big wide-angle can instantaneously be shot with video-corder omnibearing scenery, the visual field is wide, amount of image information is big, target was difficult for losing when camera and scenery were in the relative motion state, on the other hand, panorama is shot with video-corder and also be need not to use the The Cloud Terrace servo-drive system to shake bat also to have avoided the required complicated algorithm of multiple-camera image mosaic technology, saved cost, in dynamic shooting, security monitoring, maneuvering target detects, aspects such as multiple target tracking and vehicle self-navigation have wide practical use.
DSP (Digital Signal Processor) has flexibly, accurately, stablize, can repeat, volume is little, low in energy consumption, especially programmability and be easy to realize characteristics such as self-adaptive processing, having special advantages aspect high speed mathematical operations such as data, voice, video signal and the processing in real time, be the desirable hardware platform that carries out product customization and development.But current some the newer image processing algorithms of realization demand carrying out energetically applied research and exploitation urgently as still belong to the technical barrier of industry based on the complicated algorithms such as multi-valued targets tracking of Probability in DSP.The present invention considers in the realtime graphic processing procedure, work of treatment in early stage of image has that data volume is big, occupying system resources is many, processing speed requires high, the relative characteristic of simple of algorithm, adopt field programmable gate array (Field Program Gate Array is called for short FPGA) to finish in view of the above as digital picture pre-service work such as digital filtering, figure image intensifying, image segmentation and threshold process.And that DSP is responsible for data volume is less relatively, but the Flame Image Process task of algorithm, computing formula and logic strategy relative complex.Test shows that the multi-valued targets tracking means of the FPGA+DSP that the division of labor designs based on the above-mentioned functions module has been obtained good tracking effect.
The target following theory needs present information treatment technologies such as integrated use Flame Image Process, pattern-recognition, random statistical, estimation theory, optimization algorithm.Related problem is the advanced problems of multidisciplinary intersection, is the current focus of research in the world.Common method for tracking target is influenced greatly by light, when object is blocked or background complexity, background and target move etc. under the situation lose objects easily simultaneously.The present invention adopts the Monte Carlo simulation method by imparametrization to realize that recursion Bayes filtering predicts, follows the tracks of the target location, has non-linearly, and being subjected to illumination and blocking influences for a short time, expands advantages such as a plurality of targets easily.
[summary of the invention]:
The present invention seeks to overcome the prior art above shortcomings, a kind of FPGA+DSP embedded multi-valued targets threshold classification tracking means based on big wide-angle lens head breadth visual field large scene image is provided.Make this device can realize panoramic vision image acquisition, multi-valued targets threshold Classification and Identification, based on the image multiple goal of probabilistic forecasting model intersect follow the tracks of, image processing function such as effective target verification in the tracking gate.
The general frame of the FPGA+DSP embedded multi-valued targets threshold classification tracking means based on big wide-angle lens provided by the invention is:
The super large wide angle picture acquisition system of controlled resolution is made of big wide-angle lens and colour plane battle array CMOS chip able to programme, is used to gather multiple goal digital picture under the large scene.
On-site programmable gate array FPGA: be connected with the CMOS chip in the super large wide angle picture acquisition system of controlled resolution, can adjust control multiple goal digital image acquisition resolution sizes by programming; Be subjected to the control of DSP, FPGA inside IP kernel processing module reconfigured and makes up, to realize pre-service computings such as digital filtering, figure image intensifying according to its steering order; The connecting interface of CMOS and sram chip is provided, the great amount of images data in real time is transferred among the SRAM.
SRAM data-carrier store: be connected with on-site programmable gate array FPGA is two-way, be subjected to the DSP program scheduler, be used for temporary multi-valued targets digital picture.
Digital signal processor DSP: DSP is as master controller, be responsible for control FPGA collection, pretreatment image, interface and data communication task with external unit, finish the core algorithm of whole device simultaneously, comprise the multi-valued targets threshold classification and identification algorithm, intersect effective target checking algorithm in track algorithm, the tracking gate based on the image multiple goal of probabilistic forecasting model.System was loaded into program in the internal memory of DSP from FLASH when device powered on, start-up system work.
The FLASH program storage: be used for depositing the DSP handling procedure, its amount of capacity is the 32M byte, and address realm is from 0X90000000-0X92000000.
The data communication module: the FPGA+DSP embedded multi-valued targets tracking means based on big wide-angle lens is supported two kinds of communication modes: network communication and RS232 communication.The ICP/IP protocol of standard is adopted in network communication, is mainly used to transmit the output of multiple target tracking image result.RS232 works in the hardware flow control mode, to increase reliability of data transmission and stability.
Main software functional module related algorithm is
Multi-valued targets threshold Classification and Identification device:
Processing procedure is: at first carry out the target identification initialization, characteristic parameters such as the color threshold of tracked multi-valued targets and ambient lighting condition of living in are demarcated, then characteristic parameters such as the color threshold of the tracked multi-valued targets of DSP foundation and brightness carry out whole audience search to image, determine the center position of different target until system, intersect the prior probability of track algorithm with this as multiple goal at last.
Image multiple goal intersection tracker based on the probabilistic forecasting model:
This multiple goal is intersected track algorithm and be characterised in that the image multiple goal based on the probabilistic forecasting model that realizes intersects the Software Module Design thinking of tracker in DSP.Each step that the method that this multiple goal intersection track algorithm relates to a kind of hybrid cross computing of innovation is promptly decomposed in the probabilistic forecasting calculating process carries out the different target intersection respectively.These steps mainly comprise catch priori probability information (as calculating the target area color histogram), the initialization of intended particle collection, improved, calculate particle weights, weights standardization, particle resampling, target of prediction current location etc.The cross program framework of original creation makes above step carry out synchronous processing to each different target, concurrent operation, thus realized that multiple goal follows the tracks of real-time, improve processing speed greatly and followed the tracks of efficient.
Effective target checker in the tracking gate:
Processing procedure is: generate tracking gate in each target's center's point position that track algorithm dopes respectively, the big I of tracking gate is carried out program setting in advance according to the tracked target size.Carrying out the object feature value verification in tracking gate judges as each pixel being carried out color threshold, the pixel sum that meets the color of object threshold value in the tracking gate is compared with predefined ρ value, follow the tracks of effectively greater than ρ value proof, system carries out tracking results output and continues to follow the tracks of; If then follow the tracks of failure less than the ρ value, system need carry out the multi-valued targets threshold Classification and Identification of whole audience image again.The effective target checker carries out secondary checking quickly and efficiently to tracking results in the tracking gate of original creation, and the controlled target tracing process has improved reliability of system operation greatly.
Advantage of the present invention and good effect:
In sum, compare with algorithm with other tracking means, the present invention has the following advantages:
● assist colour plane battle array CMOS chip able to programme based on the overall view visual system that big wide-angle lens is set up, reach as high as 3,000,000 pixels, can be by to the programming adjustment of FPGA with control big wide angle picture resolution according to the difference of use occasion, applied environment, with high-quality target image under the optimized image scope that obtains multiple target tracking and the large scene.Optimize viewfinder range and resolution at different application, to avoid because the motion of tracked target exceeds the track rejection that the camera lens angular field of view causes and being difficult to of causing owing to target resolution is low discerned.
● high-performance FPGA+DSP embedded system fully guarantees the stability and the low-power consumption of real-time, accuracy and the device work of tracking.Be beneficial to airborne installation simultaneously, flexible configurations and maintaining easily.
● have configuration of the FPGA of novelty and dsp software module and function allocation pattern, given full play to that system is soft, hardware advantage and characteristics separately, rationally distributed, optimized device performance.
● the multi-valued targets threshold classification and identification algorithm has been considered the changeability and the complicacy of process object, adopt non-single criterion to carry out multi-threshold and estimate and target's feature-extraction, developed multivalue image partitioning algorithm and key words sorting method for registering based on color, brightness etc.Simply, practical, effective.
● adopt a kind of improved image multiple goal intersection track algorithm based on the probabilistic forecasting model, in single DSP, created a kind of cross program framework, each target is carried out synchronous processing, thereby the multi-track module is worked simultaneously to become a reality, improve processing speed and followed the tracks of efficient, realized the multiple goal real-time follow-up in every two field picture.
● create effective target checker controlled target tracing process in the tracking gate, improved reliability of system operation.Tracking results is carried out secondary checking quickly and efficiently, guarantee to catch current tracked target information based on image multiple goal intersection tracker very first time when tracking is lost of probabilistic forecasting model.
● the tracking results way of output flexibly, output image and current tracked target image coordinate simultaneously.
● rely on embedded FPGA+DSP platform, constitute the high multi-valued targets tracking means of complete, software and hardware combining, a real-time integrated level of cover.With the embedded hardware platform is that the big visual field multi-valued targets threshold classification tracker of support has wide application prospects at aspects such as dynamically shooting, security monitoring, maneuvering target detection, multiple target tracking and vehicle self-navigations.Especially have great advantage to making up volume airborne Target Tracking System little, low in energy consumption.
[description of drawings]:
Fig. 1 is a system hardware block diagram of the present invention;
Fig. 2 is a system program block diagram of the present invention.
[embodiment]:
Embodiment 1:
Following the tracks of the twin color ball navigation mark with the mobile robot, to carry out independent navigation be that example illustrates step and the method for this device when implementing identification and follow the tracks of the top navigation mark of two kinds of different colours.
Classify tracking means as shown in Figure 1 based on the FPGA+DSP multi-valued targets threshold of big wide-angle lens, comprise that mainly big wide-angle lens and colour plane battle array CMOS chip able to programme constitute the super large wide angle picture acquisition system of controlled resolution; On-site programmable gate array FPGA, the resolution sizes of programming Control images acquired, and image is carried out Digital High Pass Filter, image denoising handle, the image after handling is stored among the external data memory SRAM.DSP is a master controller, finishes the core algorithm of whole device.FLASH is as program storage: the handling procedure that is used for depositing DSP.The data communication module is connected with DSP is two-way, the communication function of implement device.
Multi-valued targets threshold classification tracing process mainly is made up of four steps:
The big wide angle picture collection of controlled resolution, multi-valued targets threshold Classification and Identification, multiple goal intersect follows the tracks of effective target verification in the tracking gate.
Following practical case is understood implementation step of the present invention and effect specifically.
Step 1: set up FPGA+DSP embedded multi-valued targets threshold classification tracking means based on big wide-angle lens.Set up super large wide-angle visual pattern acquisition system with MT9T001 type CMOS chip and FE185C046HA-1 fish eye lens.Cyclone IIC35 type FPGA is 720 * 576 with the image acquisition resolution programmed settings, and image is carried out Digital High Pass Filter, image denoising processing, and the image after handling is stored among the SRAM (CY7C1021CV33-12).TMS320DM642DSP finishes the core algorithm of whole device.FLASH (39VF040) is used for depositing the DSP handling procedure.Serial ports and network communication chip adopt 16c752 and WJLXT971A respectively.
Step 2: multi-valued targets threshold Classification and Identification.In present case, be that the target feature vector element carries out the multi-valued targets threshold discriminator to the twin color ball navigation mark with the color.For assurance and surrounding environment and target to the contrast between the color, the target surface color is set to yellow and blueness respectively.In system initialisation phase, at first outdoor light is detected, adjust FE185C046HA-1 wide-angle lens aperture; The color that yellow, blue target are moved to respectively in the image is demarcated the zone automatically then, system reads correlated variables numerical value automatically, carrying out average and variance calculates, obtain blue Y, Cb respectively, Cr threshold value limited range is 85＜Y＜123,132＜Cb＜163,100＜Cr＜125, yellow Y, Cb, Cr threshold value limited range are 97＜Y＜129,69＜Cb＜103,128＜Cr＜157; DSP utilizes multi-valued targets threshold that initial phase obtains to the judgement of classifying of each pixel in the full frame scope, statistics belongs to the pixel coordinate in blue and the yellow respectively, and the mode that the pixel coordinate utilization that belongs in the scope is separately averaged, the center that obtains blue and yellow target is respectively (423,181), (249,165), and marker ligand to output (B1, Y1).Thereby realize multiobject Classification and Identification, for step 3 provides prior probability ready.
Step 3: the multiple goal intersection is followed the tracks of.On the basis of having determined blue and yellow target's center position, DSP adopts a kind of real-time follow-up of realizing multi-valued targets based on the image multiple goal intersection track algorithm of probabilistic forecasting model, specifically comprise catch priori probability information (calculate target area color histogram), the initialization of intended particle collection, improved, calculate particle weights, weights standardization, resampling, seven steps of target of prediction current location: the first step, catch priori probability information.Need calculate color histogram respectively with image in blue and yellow target's center position 30 * 30 scopes, designed a kind of color histogram that comprises Y, U, V information simultaneously, concrete grammar is: the span 0～255 of Y, U, V component is divided into 32 sections continuously, every section size is 8, the span of three components amounts to 96 (32 * 3) section, and promptly the color histogram abscissa axis amounts to 96 coordinate points.Begin 0～32 for the span of Y from initial point, 33～64 is the span of U, 65～96 is the span of V, axis of ordinates represents to belong in yellow and blue 30 x of target's center, 30 scopes number of pixels of different Y, U, V span, utilize N number group to store the color histogram information of N target respectively, N=2 in present case, promptly the tracked target number is 2.Second step, the initialization of intended particle collection.When first two field picture arrives, in 15 * 15 scopes of the center of two targets, produce 150 30 * 30 square frame at random respectively, i.e. particle, and add up the color histogram of each particle according to the mode of the first step; The 3rd step, improved.When next two field picture arrives, according to each particle position in the previous frame (x, y), move at random new position (x+Range * random, y+Range * random), Range is a moving range, gets 15 in the present embodiment, random is-1～1 random number.In the 4th step, calculate the particle weights.Utilize
Calculate the weights of each particle, x is the multi-valued targets color histogram in the formula, and Zm is the color histogram of corresponding particle, and count is a particle number; The 5th step, the weights standardization.With the weights of the weights of each particle and all particles be divided by, make the span of each particle weights become between 0～1; In the 6th step, particle resamples.If the weights of particle are less than the threshold alpha (α=average weights * K of all particles, the particle that the big more participation particle of K value resamples is many more, K=0.2 in the present embodiment), represent that then this particle represents the probability of target location very low, need resampling can participate in next step target prediction to guarantee particle as much as possible.With reference to the method in the 3rd step, these particles are moved a position at random produce new particle, and calculate weights; The 7th step, the target of prediction current location.Setting threshold β (β=average weights * the K ' of all particles, the particle of the more little participation prediction of K ' value is many more, K ' in the present embodiment=0.8) add up respectively that weights are greater than the particle of β in two targets, the mean value of these particle weights is the predicted position of two targets.
Step 4: effective target verification in the tracking gate.In order to verify the correctness of tracking results, at two target location dot generation tracking gate 1, tracking gates 2 that dope, the tracking gate size is set at 50 x 50 respectively.In tracking gate 1, utilize color threshold 85＜Y＜123,132＜Cb＜163,100＜Cr＜125 of target 1 to add up the number of pixels that meets the demands, in tracking gate 2, utilize color threshold 97＜Y＜129,69＜Cb＜103,128＜Cr＜157 of target 2 to add up the number of pixels that meets the demands, if number of pixels is less than certain threshold value μ, then be judged to be and follow the tracks of failure, need carry out the detection of multi-valued targets again, restart the multi-valued targets tracking module; If number of pixels is more than or equal to threshold value μ then judge and follow the tracks of successfully, the output tracking result, the μ value is 100 in the present case.
Claims (4)
1, a kind of FPGA+DSP embedded multi-valued targets threshold classification tracking means based on big wide-angle lens is characterized in that the general frame of this device comprises:
The super large wide angle picture acquisition system of controlled resolution: constitute by big wide-angle lens and colour plane battle array CMOS chip able to programme, be used to gather multiple goal digital picture under the large scene;
On-site programmable gate array FPGA: be connected with the CMOS chip in the big wide angle picture acquisition system of controlled resolution, by the acquisition resolution size of programmed controlled system multiple goal digital picture; Be subjected to the control of DSP, FPGA inside IP kernel processing module reconfigured and makes up, to realize pre-service computings such as digital filtering, figure image intensifying according to its steering order; The connecting interface of CMOS and sram chip is provided, the great amount of images data in real time is transferred among the SRAM;
SRAM data-carrier store: be connected with on-site programmable gate array FPGA is two-way, be subjected to the DSP program scheduler, be used for temporary multi-valued targets digital picture;
Digital signal processor DSP: DSP is as master controller, be responsible for control FPGA collection, pretreatment image, interface and data communication task with external unit, finish the core algorithm of whole device simultaneously, comprise the multi-valued targets threshold classification and identification algorithm, intersect effective target checking algorithm in track algorithm, the tracking gate based on the image multiple goal of probabilistic forecasting model; System was loaded into program in the internal memory of DSP from FLASH when device powered on, start-up system work;
The FLASH program storage: be used for depositing the DSP handling procedure, its amount of capacity is the 32M byte, and address realm is from 0X90000000-0X92000000;
The data communication module: the FPGA+DSP embedded multi-valued targets tracking means based on big wide-angle lens is supported two kinds of communication modes: network communication and RS232 communication; The ICP/IP protocol of standard is adopted in network communication, is mainly used to transmit the output of multiple target tracking image result; RS232 works in the hardware flow control mode, to increase reliability of data transmission and stability.
2, device according to claim 1 is characterized in that the Software Module Design of the multi-valued targets threshold Classification and Identification device realized in DSP; The handling procedure of this module is: at first carry out the target identification initialization, characteristic parameters such as the color threshold of tracked multi-valued targets and ambient lighting condition of living in are demarcated, then DSP carries out whole audience search according to the color threshold of tracked multi-valued targets and the characteristic parameter of brightness to image, Classification and Identification, the mark registration, calculate the center position of related objective parameter such as area, different target, and with this basis as multiple goal intersection track algorithm.
3, device according to claim 1, the image multiple goal based on the probabilistic forecasting model that it is characterized in that realizing in DSP intersect the Software Module Design of tracker; The multiple goal intersection track algorithm that this tracker adopts relates to a kind of method of hybrid cross computing of innovation, and each step of promptly decomposing in the probabilistic forecasting calculating process carries out the different target intersection respectively; These steps mainly comprise catch priori probability information, the initialization of intended particle collection, improved, calculate that particle weights, weights standardization, particle resample, the target of prediction current location; The cross program framework of original creation makes above step carry out synchronous processing to each different target, concurrent operation, thus realized that multiple goal follows the tracks of real-time, improve processing speed greatly and followed the tracks of efficient.
4, device according to claim 1, the Software Module Design of effective target checker in the tracking gate that it is characterized in that realizing in DSP is come the controlled target tracing process; Handling procedure is: generate tracking gate in each target's center's point position that track algorithm dopes respectively, the big I of tracking gate is carried out program setting in advance according to the tracked target size; Carrying out the verification of object feature value secondary in tracking gate judges as each pixel being carried out color threshold, the pixel sum that meets the color of object threshold value in the tracking gate is compared with predefined ρ value, follow the tracks of effectively greater than ρ value proof, system carries out tracking results output and continues to follow the tracks of; If then follow the tracks of failure less than the ρ value, system need carry out the multi-valued targets threshold Classification and Identification of whole audience image again.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008101547392A CN101447075B (en) | 2008-12-31 | 2008-12-31 | Wide-angle lens-based FPGA & DSP embedded multi-valued targets threshold categorization tracking device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008101547392A CN101447075B (en) | 2008-12-31 | 2008-12-31 | Wide-angle lens-based FPGA & DSP embedded multi-valued targets threshold categorization tracking device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101447075A true CN101447075A (en) | 2009-06-03 |
| CN101447075B CN101447075B (en) | 2011-09-07 |
Family
ID=40742742
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2008101547392A Expired - Fee Related CN101447075B (en) | 2008-12-31 | 2008-12-31 | Wide-angle lens-based FPGA & DSP embedded multi-valued targets threshold categorization tracking device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101447075B (en) |
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101783020A (en) * | 2010-03-04 | 2010-07-21 | 湖南大学 | Video multi-target fast tracking method based on joint probability data association |
| CN101794515A (en) * | 2010-03-29 | 2010-08-04 | 河海大学 | Target detection system and method based on covariance and binary-tree support vector machine |
| CN101944234A (en) * | 2010-07-23 | 2011-01-12 | 中国科学院研究生院 | Multi-object tracking method and device driven by characteristic trace |
| WO2011047508A1 (en) * | 2009-10-22 | 2011-04-28 | Tianjin University Of Technology | Embedded vision tracker and mobile guiding method for tracking sequential double color beacons array with extremely wide-angle lens |
| CN102055894A (en) * | 2010-11-23 | 2011-05-11 | 无锡市博阳精密机械制造有限公司 | Modularized CCD industrial camera |
| CN102063724A (en) * | 2010-11-25 | 2011-05-18 | 四川省绵阳西南自动化研究所 | Panoramic virtual alert target relay tracking device |
| CN102158653A (en) * | 2011-05-03 | 2011-08-17 | 东华大学 | Device and method for acquiring digital image with high dynamic range in real time |
| CN102238368A (en) * | 2010-04-28 | 2011-11-09 | 长春博鸿电子科技公司（普通合伙） | Intelligent multimode multi-view integrated camera |
| CN103281518A (en) * | 2013-05-30 | 2013-09-04 | 中国科学院长春光学精密机械与物理研究所 | Multifunctional networking all-weather intelligent video monitoring system |
| CN103456171A (en) * | 2013-09-04 | 2013-12-18 | 北京英泰智软件技术发展有限公司 | Vehicle flow detection system and method based on fish-eye lens and image correction method |
| CN103716508A (en) * | 2013-12-17 | 2014-04-09 | 重庆凯泽科技有限公司 | DSP-based video image processing system |
| CN103729960A (en) * | 2012-10-13 | 2014-04-16 | 成都进界科技有限公司 | Video monitoring alarm system based on FPGA |
| CN103888751A (en) * | 2014-03-12 | 2014-06-25 | 天津理工大学 | Embedded type panoramic three-dimensional spherical visual image acquisition system based on DSP |
| CN104735298A (en) * | 2013-12-24 | 2015-06-24 | 中国科学院沈阳自动化研究所 | Video target tracking master-slave standby system and method |
| CN105005686A (en) * | 2015-07-02 | 2015-10-28 | 北京智能综电信息技术有限责任公司 | Probability prediction type target tracking method |
| CN105847650A (en) * | 2016-05-20 | 2016-08-10 | 北京科旭威尔科技股份有限公司 | Intelligent control system having target locking and tracking function |
| CN107025659A (en) * | 2017-04-11 | 2017-08-08 | 西安理工大学 | The panorama method for tracking target mapped based on unit sphere coordinate |
| CN108122026A (en) * | 2017-12-19 | 2018-06-05 | 中国人民解放军空军工程大学 | The accurate tracking of attack vehicle holder |
| CN108319918A (en) * | 2018-02-05 | 2018-07-24 | 中国科学院长春光学精密机械与物理研究所 | A kind of Embedded Trace device and the method for tracking target applied to Embedded Trace device |
| CN110869936A (en) * | 2017-06-06 | 2020-03-06 | 智加科技公司 | Method and system for distributed learning and adaptation in autonomous vehicles |
| US12039445B2 (en) | 2017-06-06 | 2024-07-16 | Plusai, Inc. | Method and system for on-the-fly object labeling via cross modality validation in autonomous driving vehicles |
-
2008
- 2008-12-31 CN CN2008101547392A patent/CN101447075B/en not_active Expired - Fee Related
Cited By (31)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2011047508A1 (en) * | 2009-10-22 | 2011-04-28 | Tianjin University Of Technology | Embedded vision tracker and mobile guiding method for tracking sequential double color beacons array with extremely wide-angle lens |
| CN101783020A (en) * | 2010-03-04 | 2010-07-21 | 湖南大学 | Video multi-target fast tracking method based on joint probability data association |
| CN101783020B (en) * | 2010-03-04 | 2011-08-17 | 湖南大学 | Video multi-target fast tracking method based on joint probability data association |
| CN101794515A (en) * | 2010-03-29 | 2010-08-04 | 河海大学 | Target detection system and method based on covariance and binary-tree support vector machine |
| CN101794515B (en) * | 2010-03-29 | 2012-01-04 | 河海大学 | Target detection system and method based on covariance and binary-tree support vector machine |
| CN102238368A (en) * | 2010-04-28 | 2011-11-09 | 长春博鸿电子科技公司（普通合伙） | Intelligent multimode multi-view integrated camera |
| CN101944234A (en) * | 2010-07-23 | 2011-01-12 | 中国科学院研究生院 | Multi-object tracking method and device driven by characteristic trace |
| CN101944234B (en) * | 2010-07-23 | 2012-07-25 | 中国科学院研究生院 | Multi-object tracking method and device driven by characteristic trace |
| CN102055894A (en) * | 2010-11-23 | 2011-05-11 | 无锡市博阳精密机械制造有限公司 | Modularized CCD industrial camera |
| CN102063724A (en) * | 2010-11-25 | 2011-05-18 | 四川省绵阳西南自动化研究所 | Panoramic virtual alert target relay tracking device |
| CN102158653A (en) * | 2011-05-03 | 2011-08-17 | 东华大学 | Device and method for acquiring digital image with high dynamic range in real time |
| CN102158653B (en) * | 2011-05-03 | 2013-01-16 | 东华大学 | Device and method for acquiring digital image with high dynamic range in real time |
| CN103729960A (en) * | 2012-10-13 | 2014-04-16 | 成都进界科技有限公司 | Video monitoring alarm system based on FPGA |
| CN103281518A (en) * | 2013-05-30 | 2013-09-04 | 中国科学院长春光学精密机械与物理研究所 | Multifunctional networking all-weather intelligent video monitoring system |
| CN103281518B (en) * | 2013-05-30 | 2016-01-13 | 中国科学院长春光学精密机械与物理研究所 | A kind of multi-network all-weather intelligent video supervisory control system |
| CN103456171A (en) * | 2013-09-04 | 2013-12-18 | 北京英泰智软件技术发展有限公司 | Vehicle flow detection system and method based on fish-eye lens and image correction method |
| CN103456171B (en) * | 2013-09-04 | 2016-04-06 | 北京英泰智软件技术发展有限公司 | A kind of based on fish-eye vehicle flow detection system, method and method for correcting image |
| CN103716508A (en) * | 2013-12-17 | 2014-04-09 | 重庆凯泽科技有限公司 | DSP-based video image processing system |
| CN104735298A (en) * | 2013-12-24 | 2015-06-24 | 中国科学院沈阳自动化研究所 | Video target tracking master-slave standby system and method |
| CN103888751A (en) * | 2014-03-12 | 2014-06-25 | 天津理工大学 | Embedded type panoramic three-dimensional spherical visual image acquisition system based on DSP |
| CN105005686B (en) * | 2015-07-02 | 2017-10-24 | 北京智能综电信息技术有限责任公司 | A kind of method for tracking target of probabilistic forecasting type |
| CN105005686A (en) * | 2015-07-02 | 2015-10-28 | 北京智能综电信息技术有限责任公司 | Probability prediction type target tracking method |
| CN105847650A (en) * | 2016-05-20 | 2016-08-10 | 北京科旭威尔科技股份有限公司 | Intelligent control system having target locking and tracking function |
| CN107025659B (en) * | 2017-04-11 | 2020-03-31 | 西安理工大学 | Panoramic target tracking method based on unit spherical coordinate mapping |
| CN107025659A (en) * | 2017-04-11 | 2017-08-08 | 西安理工大学 | The panorama method for tracking target mapped based on unit sphere coordinate |
| CN110869936A (en) * | 2017-06-06 | 2020-03-06 | 智加科技公司 | Method and system for distributed learning and adaptation in autonomous vehicles |
| US12039445B2 (en) | 2017-06-06 | 2024-07-16 | Plusai, Inc. | Method and system for on-the-fly object labeling via cross modality validation in autonomous driving vehicles |
| US12093821B2 (en) | 2017-06-06 | 2024-09-17 | Plusai, Inc. | Method and system for closed loop perception in autonomous driving vehicles |
| CN108122026A (en) * | 2017-12-19 | 2018-06-05 | 中国人民解放军空军工程大学 | The accurate tracking of attack vehicle holder |
| CN108319918A (en) * | 2018-02-05 | 2018-07-24 | 中国科学院长春光学精密机械与物理研究所 | A kind of Embedded Trace device and the method for tracking target applied to Embedded Trace device |
| CN108319918B (en) * | 2018-02-05 | 2022-07-08 | 中国科学院长春光学精密机械与物理研究所 | Embedded tracker and target tracking method applied to same |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101447075B (en) | 2011-09-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101447075B (en) | Wide-angle lens-based FPGA & DSP embedded multi-valued targets threshold categorization tracking device | |
| CN109784306B (en) | Intelligent parking management method and system based on deep learning | |
| CN101561270B (en) | Embedded omnidirectional ball vision object detection and mobile monitoring system and method | |
| CN100573388C (en) | The robot control method of real-time color auto acquisition and robot | |
| Peng et al. | Drone-based vacant parking space detection | |
| CN106027931A (en) | Video recording method and server | |
| CN102447835A (en) | Non-blind area multi-target cooperative tracking method and system | |
| Kang et al. | Application of one-stage instance segmentation with weather conditions in surveillance cameras at construction sites | |
| CN112785628A (en) | Track prediction method and system based on panoramic view angle detection and tracking | |
| CN109657639A (en) | A kind of Situation Awareness System and method based on panoramic vision | |
| Kyrkou | C 3 Net: end-to-end deep learning for efficient real-time visual active camera control | |
| Shan et al. | LMD-TShip⋆: vision based large-scale maritime ship tracking benchmark for autonomous navigation applications | |
| CN113033470B (en) | Light-weight target detection method | |
| CN118229085A (en) | Intelligent park energy management risk visual management system based on attention prediction mechanism | |
| Fehr et al. | Counting people in groups | |
| CN110008888A (en) | Comprehensive characteristics object detection method and system in intelligent monitoring network | |
| Wang et al. | Target detection for construction machinery based on deep learning and multisource data fusion | |
| CN112307943B (en) | Water area man-boat target detection method, system, terminal and medium | |
| CN103839278A (en) | Foreground detecting method and device | |
| Benet et al. | Embedded low-level video processing for surveillance purposes | |
| CN113052118A (en) | Method, system, device, processor and storage medium for realizing scene change video analysis and detection based on high-speed dome camera | |
| Zhang et al. | Optimization for 3D model-based multi-camera deployment | |
| Hu et al. | Gray spot detection in surveillance video using convolutional neural network | |
| Gregor et al. | Design and implementation of a counting and differentiation system for vehicles through video processing | |
| CN112367507A (en) | Full-time-space video enhancement management and control system based on 3D live-action model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20110907 Termination date: 20121231 |