CN110532943A - The navigation channel state analysis method combined frame by frame based on Camshift algorithm with image - Google Patents
The navigation channel state analysis method combined frame by frame based on Camshift algorithm with image Download PDFInfo
- Publication number
- CN110532943A CN110532943A CN201910802659.1A CN201910802659A CN110532943A CN 110532943 A CN110532943 A CN 110532943A CN 201910802659 A CN201910802659 A CN 201910802659A CN 110532943 A CN110532943 A CN 110532943A
- Authority
- CN
- China
- Prior art keywords
- candidate region
- navigation channel
- image
- frame
- mass center
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/24—Aligning, centring, orientation detection or correction of the image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/62—Extraction of image or video features relating to a temporal dimension, e.g. time-based feature extraction; Pattern tracking
Abstract
The invention proposes a kind of navigation channel state analysis methods combined frame by frame based on Camshift algorithm with image, it the steps include: the navigation channel state video for acquiring ship by boat-carrying camera first, label target region and mass center and candidate region and centre coordinate in the image of initial frame navigation channel are set, candidate family is established;Secondly, navigation channel image is obtained the color histogram of navigation channel image by color space conversion to hsv color space;It recycles back projection that candidate region and target area are carried out similarity measurement, acquires most similar candidate region and its mass center;Finally, it is overlapped by mobile candidate region mass center with target area mass center, then the calculating of next frame is carried out, completes the analysis frame by frame to navigation channel image.The present invention realizes the synthesis dynamic analysis of changeable navigation channel information collected to sensor using Camshift algorithm, effectively carries out dynamic navigation channel state analysis, improves the timeliness and accuracy, zero distortion ground analysis data of offer of traditional analysis mode.
Description
Technical field
The present invention relates to the technical fields of navigation channel state, particularly relate to one kind and are tied frame by frame based on Camshift algorithm with image
The navigation channel state analysis method of conjunction.
Background technique
It is transported as unmanned ship convenient and efficient and intelligent and high-efficiency in Modern Traffic carrier, in occupation of inlet and outlet goods
Very big specific gravity in object means of transportation, but hauling operation midchannel will appear various unknown problems, including peak period
Ship blockage problem, submerged reef wait many navigation channel states for being unfavorable for operation, because of unmanned ship, so will appear discovery not
In time, impolitic situation is handled, immeasurable consequence is easily caused, causes serious disaster.
If there is very big drawback when directly with sensor or some perception devices come processing environment data, i.e., can not
The changeable environment data information of acquisition is fed back in real time, does not have timeliness.General processing meeting when changeable of navigation channel state
There is big probability misjudgment phenomenon, unmanned ship is caused to receive the sail information of mistake when running, does not have accuracy.Especially occur
It the use of sensor and common processing method is directly that cannot achieve the ship high to similarity to compile when more ship blockage problems
Team carries out identification perception, so will cause serious shock accident.
Summary of the invention
For the navigation channel status information of the existing unmanned ship technical problem low there are poor in timeliness, accuracy, the present invention
A kind of navigation channel state analysis method combined frame by frame based on Camshift algorithm with image is proposed, realization collects sensor
Changeable navigation channel information synthesis dynamic analysis, effectively carry out dynamic navigation channel state analysis, improve traditional analysis mode
Timeliness and accuracy provide zero distortion ground analysis data.
The technical scheme of the present invention is realized as follows:
A kind of navigation channel state analysis method combined frame by frame based on Camshift algorithm with image, its step are as follows:
S1, the navigation channel state video that ship is acquired by boat-carrying camera obtain every frame navigation channel image, and navigate in initial frame
The centre coordinate of candidate region and candidate region is arranged in the mass center in label target region and target area in road image, establishes and waits
Modeling type;
S2, the navigation channel image in step S1 is transformed into hsv color space, obtains the color histogram of navigation channel image;
S3, image I is obtained to the color histogram progress back projection of the obtained navigation channel image of step S2, it will be in image I
Candidate region and target area be uniformly divided into the equal infinitesimal sections m respectively, recycle in Camshift algorithm
CalcBackProject function in image I candidate region and target area handle, respectively obtain the general of candidate region
Rate density puWith the probability density q of target areau, wherein u=1,2,3 ..., m;
S4, the probability density p according to the candidate region in step S3uWith the probability density q of target areauCalculate candidate regions
Similarity factor between domain and target area, and similarity factor is ranked up using the method for bubble sort, retain similar system
The maximum candidate region P of number, calculates the mass center of candidate region P;
Whether S5, the mass center for judging candidate region P are overlapped with the mass center of target area, if so, step S8 is executed, otherwise,
Execute step S6;
S6, the mass center of candidate region P is moved at the mass center of target area, and judges the shifting of the mass center of candidate region P
Whether dynamic distance is greater than threshold value T, if so, executing step S8, otherwise, executes step S7;
S7, the mass center for updating candidate region P, repeat step S5;
S8, using the mass center of candidate region P as the target centroid of next frame navigation channel image, and calculate present frame navigation channel shadow
As corresponding undistorted new coordinate position.
The image size of every frame navigation channel image in the step S1 is M*N, and the size of target area and candidate region is a*
B, the mass center of target area are f0(x'0,y'0), the centre coordinate of candidate region is (x0,y0),
Candidate region includes n pixel;The method for building up of the candidate region model are as follows: by the position number of pixel in candidate region
Group { ziIndicate:Wherein, i=1,2 ..., n, (xi,yi) be pixel coordinate.
According to the probability density p of candidate region in the step S4uWith the probability density q of target areauCalculate candidate region and
Similarity factor between target areaWherein,
δ[b(zi)-u] it is the ratio that pixel accounts for infinitesimal section in the histogram of target area in candidate region, C is normalization coefficient, and n is to wait
The number of the pixel of favored area, KEIt (x) is the kernel function of the target area of navigation channel image,
Function k () is non-negative, non-increasing, segmentation and continuous function, and
The mass center of the candidate region P are as follows: ft(xt,yt), wherein abscissa xt=M01/M00, ordinate yt=M10/M00,For the zeroth order square of candidate region P,With
It is the first moment of candidate region P.
The kernel function KE(x) window size is h*h, and the threshold value T in the step S6 is 1/2h.
The update method of the mass center of candidate region P in the step S7 are as follows:
Wherein, g (x)=- K'E(x), the independent variable of x representation space Kernel Function,For weight coefficient, K'E
It (x) is kernel function KE(x) differential,For the probability density of candidate region.
The corresponding undistorted new coordinate position of present frame navigation channel image in the step S8 are as follows:
It is that the technical program can generate the utility model has the advantages that
The center-of-mass coordinate of target area and the center of candidate region are calculated to a frame image based on Camshift algorithm
Position coordinates overcome the inflexible problem of existing image processing means, have more flexibility, high fault tolerance in contrast;
It drives region of search to move validity feature region the movement of candidate regions center pixel, realizes more effectively special
Probability distribution is levied to region of search;The disadvantage that sensor acquisition image can not be eliminated there are deviation is overcome, is had stronger
Anti-interference ability is on a frame-by-frame basis inputted based on the method that image is analyzed frame by frame using by the navigation channel state image of real-time change
Into tracking cipher rounds, the problem of unqualified image frame of analysis leads to erroneous judgement is effectively eliminated, because navigation channel state is
It is changeable, but existing technological means only simply identifies the still image that sensor obtains at present, in this way since, answer
It is difficult to realize when using in unmanned ship, increases tracking cipher rounds and greatly reduce the shadow because of still image processing
Ringing has stronger dynamic;The present invention highlights dynamic, real-time, continuity, overcomes in conventional method and does not have in fact
Shi Xing, successional problem reach accurate and dynamic following analysis level.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
It obtains other drawings based on these drawings.
Fig. 1 is flow chart of the invention.
Fig. 2 is feature pixel probability density distribution preset model figure of the invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, those of ordinary skill in the art are obtained every other under that premise of not paying creative labor
Embodiment shall fall within the protection scope of the present invention.
As shown in Figure 1, the embodiment of the invention provides a kind of navigation channels combined frame by frame based on Camshift algorithm with image
State analysis method, the specific steps are as follows:
S1, the navigation channel state video that ship is acquired by boat-carrying camera obtain every frame navigation channel image, and navigate in initial frame
The mass center in artificial label target region and target area in road image is arranged the centre coordinate of candidate region and candidate region, builds
Vertical candidate family calculates generation color histogram for next step and provides input quantity.
The image size of every frame navigation channel image is M*N, and the size of target area and candidate region is a*b, mesh
The mass center for marking region is f0(x'0,y'0), the centre coordinate of candidate region is (x0,y0),
Candidate region includes n pixel.
The method for building up of the candidate region model are as follows: by the position of pixel in candidate region array { ziIndicate:Wherein, i=1,2 ..., n, (xi,yi) be pixel coordinate.
S2, the navigation channel image in step S1 is transformed into hsv color space, obtains the color histogram of navigation channel image;RGB
Color space is more sensitive to the perception of light condition, in order to eliminate influence of the factor to tracking effect, by each frame navigation channel
Image is converted into hsv color space, screens and extracts tone H component to apply Camshift algorithm pair in subsequent step
The hue histogram in the frame region calculates.The color histogram that tone H is had in the dynamic image frame region of navigation channel can be obtained in this way
Figure.
S3, it is obtained using color histogram progress back projection of the Camshift algorithm to the obtained navigation channel image of step S2
Image I, by image I candidate region and target area be uniformly divided into the equal infinitesimal sections m respectively, recycle
CalcBackProject function in Camshift algorithm in image I candidate region and target area handle, respectively
Obtain the probability density p of candidate regionuWith the probability density q of target areau, wherein u=1,2,3 ..., m;
S4, the probability density p according to the candidate region in step S3uWith the probability density q of target areauCalculate candidate regions
Similarity factor between domain and target area, and similarity factor is ranked up using the method for bubble sort, retain similar system
The maximum candidate region P of number, then calculate the mass center of candidate region P.
During actual tracking, when candidate target, which occurs blocking etc., to be influenced, since the pixel value of outer layer is easy to be hidden
The influence of gear or light, so the pixel ratio edge pixel values of target area immediate vicinity are more reliable.Therefore, for all samplings
The importance of point, each sample point should be different, and remoter from central point, weight is smaller.Therefore introduce kernel function KE(x) and
Weight coefficient improves the robustness of iteration tracing algorithm and increases searching analysis ability.
As shown in Fig. 2, according to the probability density p of candidate regionuWith the probability density q of target areauCalculate candidate region and mesh
Mark the similarity factor ρ (p, q) between region:Wherein,δ
[b(zi)-u] it is the ratio that pixel accounts for infinitesimal section in the histogram of target area in candidate region, C is normalization coefficient, and n is to wait
The number of the pixel of favored area, KEIt (x) is the kernel function of the target area of navigation channel image, kernel function KE(x) window size is h*
H,Function k () is non-negative, non-increasing, segmentation and continuous function, and
In the back projection of candidate region histogram, I (α, β) is the candidate region of the back projection figure of color histogram
In the corresponding coordinate value of the pixel, the relative position of mass center in present frame is determined according to zeroth order square, first moment.Candidate region P
Mass center are as follows: ft(xt,yt), wherein abscissa xt=M01/M00, ordinate yt=M10/M00,To wait
The zeroth order square of favored area P,WithIt is the single order of candidate region P
Square.The mass center of candidate region is adjusted to the mass center of target area, the size of the zeroth order square and candidate region that are calculated is linear
Correlation, the mass center for making the mass center of candidate region follow target area in real time is mobile, remains that two hearts are overlapped.
Whether S5, the mass center for judging candidate region P are overlapped with the mass center of target area, if so, step S8 is executed, otherwise,
Execute step S6.
S6, the mass center of candidate region P is moved at the mass center of target area, and judges the shifting of the mass center of candidate region P
Whether dynamic distance is greater than threshold value T, T=1/2h, if so, executing step S8, otherwise, executes step S7.
S7, pass through and threshold value T comparison, each result successive iteration calculate, find the true mass center of current frame image
Value updates the mass center of candidate region P, repeats step S5.The update method of the mass center of candidate region P are as follows:Wherein, g (x)=- K'E(x), the independent variable of x representation space Kernel Function,For weight coefficient, K'EIt (x) is kernel function KE(x) differential,
For the probability density of candidate region, weight coefficient wiShow that candidate region is more similar to target area more greatly.
S8, using the mass center of candidate region P as the mass center of the target area of next frame navigation channel image, and calculate present frame
The corresponding undistorted new coordinate position of navigation channel image.The corresponding undistorted new coordinate position of present frame navigation channel image are as follows:
So far, it is completed based on Camshift algorithm with the navigation channel state analysis method that image is combined frame by frame.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Within mind and principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (7)
1. a kind of navigation channel state analysis method combined frame by frame based on Camshift algorithm with image, which is characterized in that its step
It is as follows:
S1, the navigation channel state video that ship is acquired by boat-carrying camera, obtain every frame navigation channel image, and in initial frame navigation channel shadow
The centre coordinate of candidate region and candidate region is arranged in the mass center of label target region and target area as in, establishes candidate mould
Type;
S2, the navigation channel image in step S1 is transformed into hsv color space, obtains the color histogram of navigation channel image;
S3, image I is obtained to the color histogram progress back projection of the obtained navigation channel image of step S2, by the time in image I
Favored area and target area are uniformly divided into m equal infinitesimal sections respectively, recycle in Camshift algorithm
CalcBackProject function in image I candidate region and target area handle, respectively obtain the general of candidate region
Rate density puWith the probability density q of target areau, wherein u=1,2,3 ..., m;
S4, the probability density p according to the candidate region in step S3uWith the probability density q of target areauCalculate candidate region and
Similarity factor between target area, and similarity factor is ranked up using the method for bubble sort, retain similarity factor most
Big candidate region P calculates the mass center of candidate region P;
Whether S5, the mass center for judging candidate region P are overlapped with the mass center of target area, if so, executing step S8, otherwise, execute
Step S6;
S6, the mass center of candidate region P is moved at the mass center of target area, and judge candidate region P mass center movement away from
From whether threshold value T is greater than, if so, executing step S8, otherwise, step S7 is executed;
S7, the mass center that candidate region P is updated according to step S6, repeat step S5;
S8, using the mass center of candidate region P as the target centroid of next frame navigation channel image, and calculate present frame navigation channel image pair
The undistorted new coordinate position answered.
2. the navigation channel state analysis method according to claim 1 combined frame by frame based on Camshift algorithm with image, feature
It is, the image size of every frame navigation channel image in the step S1 is M*N, and the size of target area and candidate region is a*b,
The mass center of target area is f0(x'0,y'0), the centre coordinate of candidate region is (x0,y0),
Candidate region includes n pixel;The method for building up of the candidate region model are as follows: by the position number of pixel in candidate region
Group { ziIndicate:Wherein, i=1,2 ..., n, (xi,yi) be pixel coordinate.
3. the navigation channel state analysis method according to claim 1 combined frame by frame based on Camshift algorithm with image, feature
It is, according to the probability density p of candidate region in the step S4uWith the probability density q of target areauCalculate candidate region and mesh
Mark the similarity factor ρ (p, q) between region:Wherein,δ
[b(zi)-u] it is the ratio that pixel accounts for infinitesimal section in the histogram of target area in candidate region, C is normalization coefficient, and n is to wait
The number of the pixel of favored area, KEIt (x) is the kernel function of the target area of navigation channel image,
Function k () is non-negative, non-increasing, segmentation and continuous function, and
4. the navigation channel state analysis method according to claim 2 combined frame by frame based on Camshift algorithm with image, special
Sign is, the mass center of the candidate region P are as follows: ft(xt,yt), wherein abscissa xt=M01/M00, ordinate yt=M10/M00,For the zeroth order square of candidate region P,With
It is the first moment of candidate region P.
5. the navigation channel state analysis method according to claim 3 combined frame by frame based on Camshift algorithm with image,
It is characterized in that, the kernel function KE(x) window size is h*h, and the threshold value T in the step S6 is 1/2h.
6. the navigation channel state analysis method according to claim 3 or 4 combined frame by frame based on Camshift algorithm with image, special
Sign is, the update method of the mass center of candidate region P in the step S7 are as follows:
Wherein, g (x)=- K'E(x), the independent variable of x representation space Kernel Function,For weight coefficient, K'E
It (x) is kernel function KE(x) differential,For the probability density of candidate region.
7. the navigation channel state analysis method according to claim 6 combined frame by frame based on Camshift algorithm with image,
It is characterized in that, the corresponding undistorted new coordinate position of the present frame navigation channel image in the step S8 are as follows:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910802659.1A CN110532943A (en) | 2019-08-28 | 2019-08-28 | The navigation channel state analysis method combined frame by frame based on Camshift algorithm with image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910802659.1A CN110532943A (en) | 2019-08-28 | 2019-08-28 | The navigation channel state analysis method combined frame by frame based on Camshift algorithm with image |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110532943A true CN110532943A (en) | 2019-12-03 |
Family
ID=68664865
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910802659.1A Pending CN110532943A (en) | 2019-08-28 | 2019-08-28 | The navigation channel state analysis method combined frame by frame based on Camshift algorithm with image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110532943A (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107818571A (en) * | 2017-12-11 | 2018-03-20 | 珠海大横琴科技发展有限公司 | Ship automatic tracking method and system based on deep learning network and average drifting |
CN108447073A (en) * | 2017-11-22 | 2018-08-24 | 西北工业大学 | It is a kind of based on DSP nobody is aerial plus by oily laser vision tracking |
CN109102523A (en) * | 2018-07-13 | 2018-12-28 | 南京理工大学 | A kind of moving object detection and tracking |
US20190065825A1 (en) * | 2017-08-23 | 2019-02-28 | National Applied Research Laboratories | Method for face searching in images |
CN109886079A (en) * | 2018-12-29 | 2019-06-14 | 杭州电子科技大学 | A kind of moving vehicles detection and tracking method |
-
2019
- 2019-08-28 CN CN201910802659.1A patent/CN110532943A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20190065825A1 (en) * | 2017-08-23 | 2019-02-28 | National Applied Research Laboratories | Method for face searching in images |
CN108447073A (en) * | 2017-11-22 | 2018-08-24 | 西北工业大学 | It is a kind of based on DSP nobody is aerial plus by oily laser vision tracking |
CN107818571A (en) * | 2017-12-11 | 2018-03-20 | 珠海大横琴科技发展有限公司 | Ship automatic tracking method and system based on deep learning network and average drifting |
CN109102523A (en) * | 2018-07-13 | 2018-12-28 | 南京理工大学 | A kind of moving object detection and tracking |
CN109886079A (en) * | 2018-12-29 | 2019-06-14 | 杭州电子科技大学 | A kind of moving vehicles detection and tracking method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108320510B (en) | Traffic information statistical method and system based on aerial video shot by unmanned aerial vehicle | |
CN106127204B (en) | A kind of multi-direction meter reading Region detection algorithms of full convolutional neural networks | |
CN110188720A (en) | A kind of object detection method and system based on convolutional neural networks | |
US20210215481A1 (en) | Method for measuring antenna downtilt angle based on multi-scale deep semantic segmentation network | |
CN104318258B (en) | Time domain fuzzy and kalman filter-based lane detection method | |
CN108875794B (en) | Image visibility detection method based on transfer learning | |
CN106909902B (en) | Remote sensing target detection method based on improved hierarchical significant model | |
CN103148881B (en) | Method and device for automatically reading pointer instrument by video frequency | |
CN107506729B (en) | Visibility detection method based on deep learning | |
CN110363122A (en) | A kind of cross-domain object detection method based on multilayer feature alignment | |
CN111079602A (en) | Vehicle fine granularity identification method and device based on multi-scale regional feature constraint | |
CN102110227B (en) | Compound method for classifying multiresolution remote sensing images based on context | |
CN109035292A (en) | Moving target detecting method and device based on deep learning | |
CN107330861B (en) | Image salient object detection method based on diffusion distance high-confidence information | |
CN106156758B (en) | A kind of tidal saltmarsh method in SAR seashore image | |
CN105957107A (en) | Pedestrian detecting and tracking method and device | |
CN113033315A (en) | Rare earth mining high-resolution image identification and positioning method | |
CN110307903A (en) | A kind of method of the contactless temperature dynamic measurement of poultry privileged site | |
CN114639064B (en) | Water level identification method and device | |
Bu et al. | A UAV photography–based detection method for defective road marking | |
CN104598914A (en) | Skin color detecting method and device | |
CN104463896A (en) | Image corner point detection method and system based on kernel similar region distribution characteristics | |
CN110008834B (en) | Steering wheel intervention detection and statistics method based on vision | |
CN110532943A (en) | The navigation channel state analysis method combined frame by frame based on Camshift algorithm with image | |
Sun et al. | Decoupled feature pyramid learning for multi-scale object detection in low-altitude remote sensing images |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191203 |