CN105261037B - A kind of moving target detecting method of adaptive complex scene - Google Patents
A kind of moving target detecting method of adaptive complex scene Download PDFInfo
- Publication number
- CN105261037B CN105261037B CN201510645189.4A CN201510645189A CN105261037B CN 105261037 B CN105261037 B CN 105261037B CN 201510645189 A CN201510645189 A CN 201510645189A CN 105261037 B CN105261037 B CN 105261037B
- Authority
- CN
- China
- Prior art keywords
- image
- background
- model
- value
- carried out
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 39
- 230000003044 adaptive effect Effects 0.000 title claims abstract description 13
- 238000005286 illumination Methods 0.000 claims abstract description 28
- 230000011218 segmentation Effects 0.000 claims abstract description 23
- 230000000877 morphologic effect Effects 0.000 claims abstract description 9
- 238000006243 chemical reaction Methods 0.000 claims abstract description 8
- 238000002372 labelling Methods 0.000 claims abstract description 4
- 238000000605 extraction Methods 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 6
- 241000526657 Microchloa Species 0.000 claims description 5
- 238000005530 etching Methods 0.000 claims description 5
- 238000003384 imaging method Methods 0.000 claims description 4
- 239000000203 mixture Substances 0.000 claims description 4
- 238000010276 construction Methods 0.000 claims description 3
- 230000003247 decreasing effect Effects 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000001514 detection method Methods 0.000 abstract description 14
- 230000001154 acute effect Effects 0.000 abstract description 2
- 239000004744 fabric Substances 0.000 description 5
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000036624 brainpower Effects 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000035772 mutation Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 239000007787 solid Substances 0.000 description 2
- 230000003542 behavioural effect Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000002156 mixing Methods 0.000 description 1
- 210000001260 vocal cord Anatomy 0.000 description 1
- 239000011800 void material Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration using local operators
- G06T5/30—Erosion or dilatation, e.g. thinning
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of moving target detecting method of adaptive complex scene, 1) illumination compensation is carried out to video image;2) background image of every frame video image is obtained using mixed Gaussian background modeling method;3) background subtraction principle is utilized to obtain the absolute difference image per frame;4) maximum entropy segmentation principle is used to obtain the gray probability model optimum segmentation threshold value of each absolutely difference image;5) binary conversion treatment is carried out to obtain foreground image to absolute difference image using optimum segmentation threshold value;6) module of different structure body is used to carry out Morphological scale-space;7) region labeling is carried out to foreground image using connected domain calibration algorithm, the moving target demarcated is locked using rectangle frame.This method has preferable motion target adaptive detection accuracy and robustness under the different complex scenes such as global illumination acute variation, background interference, relative motion, can improve the performance of target detection.
Description
Technical field
The present invention relates to video brainpower watch and control technology more particularly to a kind of moving object detection sides of adaptive complex scene
Method belongs to technical field of image processing.
Background technology
Detection for Moving Target is one of the key technology of video brainpower watch and control technical field, be target identities identification,
The basis of the follow-up studies such as tracking, behavioural analysis.Common Detection for Moving Target has optical flow method, frame differential method, background subtraction
Point-score.Wherein, optical flow method is a kind of motion conditions of the pixel of estimated sequence image in continuous interframe, since this method is only closed
The pixel of heart image is difficult to accomplish accurately to the irregular target of profile there is no pixel is associated with moving target
Positioning, and operation is complicated.Frame differential method is preferable to the applicability of scene changes, especially the scene of illumination variation, but to ring
Border noise is more sharp, and extracted target area is the "or" region of target position in front and back two frame, than realistic objective region
Greatly, if without notable movement tendency in tracking scene, target lap does not measure inspection between two frames, or detected
Target area exists compared with macroscopic-void, can not completely extract moving target.The key of background subtraction is background modeling and threshold
The selection of value, basic principle are to subtract background image using present frame, and combine threshold value to obtain motion target area.It utilizes
Traditional Gauss background modeling, average background modeling, median background modeling etc., are vulnerable to Changes in weather, illuminance abrupt variation, background perturbation
And the influence of the factors such as camera and target relative movement, fixed threshold does not have adaptability in addition, for example, threshold value selection is too low,
It is not enough to inhibit noise in image;It selects excessively high, then has ignored variation useful in image;For larger, solid colour
Moving target, it is possible to the problem of target internal generates cavity, can not completely extract moving target.Although background subtraction
In the case of stationary background and ideal scenario, target detection effect is preferable, but since actual scene is complicated, Changes in weather,
Global illumination mutation, background perturbation and the factors such as camera and target relative movement easily lead to moving object detection inaccuracy.
Invention content
For deficiencies of the prior art, the object of the present invention is to provide a kind of movements of adaptive complex scene
Object detection method.This method have under the different complex scene such as global illumination acute variation, background interference, relative motion compared with
Good motion target adaptive detection accuracy and robustness.This method can improve the property of target detection under complex scene
Can, provide more steady basis for the operation of link below.
Purpose to realize the present invention uses following technical scheme:
A kind of moving target detecting method of adaptive complex scene, steps are as follows,
1) video image is obtained, illumination compensation is carried out to video image, the influence brought to overcome global illumination to be mutated;
2) the corresponding background image of every frame video image is obtained using mixed Gaussian background modeling method;
3) the absolute difference image per frame is obtained, and carry out using background subtraction principle according to the background image of extraction
Median filter process, to slacken influence of noise;
4) use the gray probability model that maximum entropy segmentation principle obtains filtered each absolutely difference image corresponding
Optimum segmentation threshold value;
5) binary conversion treatment is carried out to filtered each absolutely difference image using corresponding optimum segmentation threshold value
To obtain foreground image;
6) on the basis of step 5) obtains foreground image, Morphological scale-space is carried out using the module of different structure body, with
The influence that small grass comes is eliminated, the cavity of componental movement target area is made up;" decussate texture " template of 3*3 cores is used first
It carries out an etching operation and then carries out expansive working twice with 5*3 cores, then once corroded to remove some small noises
Operation;
7) region labeling is carried out using the foreground image after connected domain calibration algorithm pair the 6) step Morphological scale-space, utilizes square
Shape frame locks the moving target demarcated.
Wherein, the illumination compensation of step 1) carries out as follows,
Assuming that I (t) indicates that inputted video image frame, δ indicate that two interframe allow the maximum global illumination occurred variation;First
Calculate the average pixel value of each frame sequence image of videoThen illumination compensation is carried out using following rule:
In formula, sgn () indicates sign function,Indicate the image after compensation.
Wherein, step 4) optimum segmentation threshold value acquisition methods are,
If a width size is the image I (x, y) of M*N, I (x, y) indicates the grey scale pixel value of image coordinate point (x, y), and
Gray value value range is 0- (L-1), and the filtered absolute difference image of step 3) is DF (x, y), niIndicate absolute difference diagram
The gray value of picture is the number of pixels of i, then number of pixels total amount is:piIndicate the probability of grey scale pixel value i, then:
pi=ni/ N, i=0,1,2,3 ..., L-1;
Then use segmentation candidates threshold value T that the pixel value in image is divided into two class of C0 and C1 by tonal gradation, C0 is indicated
Target object, C1 indicate background object, i.e. C0={ 0,1 ..., T }, C1={ T+1, T+2 ..., L-1 }, then corresponding to C0 and C1
Grey scale pixel value probability distribution is respectively:
C0:
C1:
In formula,L is the number of gray level;So, the entropy of C0 and C1 is expressed from the next respectively;
C0:
C1:
On the basis of gained image C0 entropys and C1 entropys, then the sum of posterior entropy H is indicated as follows:
H=H0+H1;
So, compare to obtain tonal gradation corresponding to the maximum value of entropy-discriminate function, that is, indicate based on maximum entropy algorithm
Optimum segmentation threshold value THR, is shown below,
Binary conversion treatment is carried out to filtered absolute difference image DF (x, y) using the optimum segmentation threshold value THR of acquisition,
The foreground image FI (x, y) in video is obtained, is shown below,
Wherein, step 2) using mixed Gaussian background modeling method extraction background image specific method be,
Using the gauss hybrid models of the K a certain pixel X of single gaussian probability model construction, see shown in formula (3);
Wherein, p (Xt) it is that t moment pixel value X occurstProbability, wi,tIndicate the weights of i-th of Gauss model of t moment, and
And weights and indicate Gauss model sum for 1, K, take 3-5, η (Xt,μi,t,∑i,t) indicate i-th of Gauss model of t moment,
μi,tFor mean value, ∑i,tFor covariance matrix, n representation dimensions are shown in formula (4);
Mixture Gaussian background model matching is as follows with renewal process:
Model Matching be video image current frame pixel value X and existing K Gauss model are subjected to matching comparison, if
I Gauss model meets formula (5), then it represents that current frame pixel value is matching, otherwise mismatches;
|Xt-μi,t-1| 2.5 σ of <i,t-1 (5)
If matching is unsuccessful, the mean value of video present frame is used, and set a larger variance yields, establish new Gauss
Distributed model;
The update of model is carried out according to formula (6) according to matching result;
Wherein, α indicates that video present frame is embedded into the rate of background model, referred to as learning rate, if Model Matching,
Mi,t=1, it is otherwise 0, μ and σ2It remains unchanged;
Due to ∑i,tThe smaller and big gaussian probability distributed model of weights is more likely used for approximate representation background pixel point
Cloth model, for this purpose, dividing K gaussian probability according to the sequence that the size of w/ σ values is successively decreased per the pixel value in frame image video
Cloth model sorts, and regard the distribution of preceding B gaussian probability as background, constitutes background image BI, see formula (7);
Wherein, τ is the threshold value of background model setting, τ value ranges [0.7,0.8].
Compared with the conventional method, the present invention has the advantages that:
1) video background image need not be preset.
2) it uses illumination compensation and mixed Gauss model to establish background model, can effectively overcome illuminance abrupt variation, camera
Relative motion imaging, the influence of background perturbation, to obtain more steady background image.
3) present invention introduces maximum entropy segmentation threshold, each absolutely difference image is respectively calculated and obtains (i.e. each
Absolute difference image may be different, and existing is fixed threshold, i.e., all absolute difference image threshold values are identical), in practical application
Involved different complex scene video images, fixed threshold, which does not have the problem of adaptability, to be resolved well.
4) there is preferable accuracy and robust under the different complex scenes such as illuminance abrupt variation, background interference, relative motion
Property.
Description of the drawings
The moving target detecting method overall framework figure of the adaptive complex scene of Fig. 1-present invention.
The flow chart of Fig. 2-mixed Gaussian background modelings of the present invention.
Fig. 3-step 2 schematic diagrams of the present invention.
Specific implementation mode
General thought of the present invention is:First, it is contemplated that the degree of illumination variation, introducing illumination compensation method improves illumination variation
Influence to succeeding target detection;Second, it is contemplated that the key of background subtraction is background modeling and threshold value selection, utilizes mixing
Gaussian Background modeling extraction background image utilizes the back of the body on this basis with the influence for overcoming dynamic background to detect succeeding target
Scape calculus of finite differences principle obtains absolute difference image, and introduces medium filtering and be filtered first to absolute difference image, with
Slacken the influence of noise, in addition the fixed defect of threshold value in original background calculus of finite differences, introduce maximum entropy split plot design extraction threshold value with
Just adaptive different complex scene video images;Third, it is contemplated that there are small noise and the same areas for the foreground image of acquisition
There are disconnected factors, Morphological scale-space are carried out to foreground image using the module of different structure body, to eliminate small grass
The influence come, makes up the cavity of componental movement target area.Finally, foreground object is marked using connected domain calibration algorithm, and
According to connected domain size lock motion target.
The specific technical solution of the present invention is as follows, and principle is shown in Fig. 1:
Step 1:Detection video is obtained, background model is established using illumination compensation and mixed Gauss model, to obtain more
Steady background image.Obtain the detailed process of background image:
(1) video sequence image is obtained, illumination compensation is carried out to video image first, to overcome global illumination mutation to bring
Interference.
Assuming that I (t) indicates that inputted video image frame, δ indicate that two interframe allow the maximum global illumination occurred variation.First
Calculate the average pixel value of each frame sequence image of videoThen illumination compensation is carried out using following rule:
In formula, sgn () indicates sign function,Indicate the image after compensation.
(2) on this basis, background image is extracted using mixed Gaussian background modeling method, camera phase can be suitable for
It is as shown in Figure 2 to dynamic scenes, the flows of mixed Gaussian background modeling such as movement imaging, background perturbation, Changes in weather.
Mixture Gaussian background model is a kind of extended pattern list Gauss model, can be with the probability of approximate representation any shape point
Cloth.In the model, the variation of pixel point value is treated as random process and meets Gaussian Profile in sequence of video images, utilizes K
The gauss hybrid models of a certain pixel X of a list gaussian probability model construction, are shown in shown in formula (3).
Wherein, p (Xt) it is that t moment pixel value X occurstProbability, wi,tIndicate the weights of i-th of Gauss model of t moment, and
And weights and indicate Gauss model sum for 1, K, generally take 3-5, η (Xt,μi,t,∑i,t) indicate i-th of Gaussian mode of t moment
Type, μI, tFor mean value, ∑I, tFor covariance matrix, n representation dimensions are shown in formula (4).
Mixture Gaussian background model mainly considers that matching and replacement problem, matching are as follows with renewal process:
Model Matching be video image current frame pixel value X and existing K Gauss model are subjected to matching comparison, if
I Gauss model meets formula (5), then it represents that current frame pixel value is matching, otherwise mismatches.
|Xt-μi,t-1| 2.5 σ of <i,t-1 (5)
If matching is unsuccessful, the mean value of video present frame is used, and set a larger variance yields, establish new Gauss
Distributed model.
The update of model is carried out according to following formula (6) according to matching result.
Wherein, α indicates that video present frame is embedded into the rate of background model, referred to as learning rate, if Model Matching,
Mi,t=1, it is otherwise 0, μ and σ2It remains unchanged.
Due to ∑i,tThe smaller and big gaussian probability distributed model of weights is more likely used for approximate representation background pixel point
Cloth model, for this purpose, dividing K gaussian probability according to the sequence that the size of w/ σ values is successively decreased per the pixel value in frame image video
Cloth model sorts, and regard the distribution of preceding B gaussian probability as background, constitutes background image BI, see formula (7).
Wherein, τ is the threshold value of background model setting, if the value is smaller, model will be degenerated to single gaussian probability distributed mode
Type;If the value is larger, it can indicate that more complicated background model, many experiments illustrate, τ optimum valuing ranges [0.7,
0.8]。
Step 2:Absolute difference image D (x, y) is obtained using background subtraction principle, and in the progress of absolute difference image
Value filtering processing.The basic principle of background subtraction is that present frame and background image are carried out absolute differential process, such as formula (8) institute
Show.
D (x, y)=| I (x, y)-BI (x, y) (8)
Wherein, I (x, y) indicates that video current frame image, BI (x, y) indicate the background image obtained by step 1, step
Rapid 2 schematic diagram is as shown in Figure 3.
Step 3:The optimal threshold of each absolutely difference image is adaptively obtained using maximum entropy segmentation threshold method, and is carried out
Binary conversion treatment obtains foreground image.Obtain the detailed process of foreground image:
If a width size is the image I (x, y) of M*N, I (x, y) indicates the grey scale pixel value of image coordinate point (x, y), and
Gray value value range is 0~(L-1), and the absolute difference image DF (x, y) of filtered background, n are obtained by step 2iIt indicates
The gray value of absolute difference image is the number of pixels of i, then number of pixels total amount is:piIndicate grey scale pixel value i's
Probability, then:
pi=ni/ N, i=0,1,2,3 ..., L-1 (9)
Then use segmentation candidates threshold value T that the pixel value in image is divided into 2 class C0 and C1, C0 expression mesh by tonal gradation
Object is marked, C1 indicates background object, i.e. C0={ 0,1 ..., T }, C1={ T+1, T+2 ..., L-1 }, then picture corresponding to C0 and C1
Plain gray value probability distribution is respectively:
C0:
C1:
In formula,L is the number of gray level.So, the entropy of C0 and C1 is respectively by formula (12) (13) table
Show.
C0:
C1:
On the basis of gained image C0 entropys and C1 entropys, then the sum of posterior entropy H is indicated as follows:
H=H0+H1 (14)
So, compare to obtain tonal gradation corresponding to the maximum value of entropy-discriminate function, that is, indicate based on maximum entropy algorithm
Optimal threshold THR is shown in shown in formula (15).
Binary conversion treatment is carried out to the difference image DF (x, y) obtained by step 2 using the optimal threshold THR of acquisition, is obtained
Foreground image FI (x, y) in video is shown in shown in formula (16).
Step 4:Morphological operation is carried out to the foreground image that step 3 obtains, to eliminate the influence that small grass comes, is made up
The cavity of componental movement target area.
Morphological scale-space is to solve to eliminate the influence that small grass comes, and makes up the cavity of componental movement target area.First
An etching operation is carried out with " decussate texture " template of 3*3 cores, some small noises can be removed, then carried out with 5*3 cores
Reexpansion operates, then carries out an etching operation.It is longitudinal that common pedestrian is allowed for as movement mesh using larger core
There are the number of people and trunk not to be connected to for mark object, can suitably do a little compensation.
Step 5:Connected domain calibration algorithm carries out region labeling to foreground image, and the movement demarcated is locked using rectangle frame
Target.
It can be seen from the above description that present invention mainly solves problems with:
1, background modeling problem
The present invention establishes background model using illumination compensation and mixed Gauss model, can effectively overcome global illumination prominent
Change, camera relative motion imaging, the influence of background perturbation.
2, fixed threshold problem
Present invention introduces maximum entropy segmentation threshold, each absolutely difference image respectively obtains corresponding segmentation threshold, in reality
Involved different complex scene video images are applied on border, and fixed threshold, which does not have the problem of adaptability, to be solved well
Certainly.For example, threshold value selection is too low, it is not enough to inhibit noise in image;It selects excessively high, then has ignored variation useful in image,
For larger, solid colour moving target, it is possible to generate cavity in target internal, can not completely extract moving target
The problem of.
3, there is preferable moving object detection accuracy and robustness
The present invention uses illumination compensation and mixed Gauss model to establish background model first, then utilizes background subtraction former
Reason carries out absolute difference and separately wins to obtain absolute difference image, and maximum entropy segmentation threshold method is recycled adaptively to obtain absolute difference image
Optimal threshold, and carry out binary conversion treatment obtain foreground image, then to foreground image carry out morphological operation, to eliminate small make an uproar
The influence that vocal cords come, makes up the cavity of componental movement target area, and finally, connected domain calibration algorithm carries out region to foreground image
Calibration locks the moving target demarcated using rectangle frame.This method can be in global illumination variation, background interference, opposite fortune
There is preferable moving object detection accuracy and robustness under the different complex scenes such as dynamic.
Finally illustrate, above-described embodiment is only used to illustrate the technical scheme of the present invention and unrestricted, although with reference to compared with
Good embodiment describes the invention in detail, it will be understood by those of ordinary skill in the art that, it can be to the skill of the present invention
Art scheme is modified or replaced equivalently, and without departing from the objective and range of technical solution of the present invention, should all be covered at this
In the right of invention.
Claims (3)
1. a kind of moving target detecting method of adaptive complex scene, camera relative motion imaging, it is characterised in that:Step
It is as follows,
1) video image is obtained, illumination compensation is carried out to video image, the influence brought to overcome global illumination to be mutated;
2) the corresponding background image of every frame video image is obtained using mixed Gaussian background modeling method;
3) the absolute difference image per frame is obtained, and carry out intermediate value using background subtraction principle according to the background image of extraction
It is filtered, to slacken influence of noise;
4) use the gray probability model that maximum entropy segmentation principle obtains filtered each absolutely difference image corresponding optimal
Segmentation threshold;
5) binary conversion treatment is carried out to obtain to filtered each absolutely difference image using corresponding optimum segmentation threshold value
Obtain foreground image;
6) on the basis of step 5) obtains foreground image, Morphological scale-space is carried out using the module of different structure body, to eliminate
The influence that small grass comes, makes up the cavity of componental movement target area;It is carried out first with " decussate texture " template of 3*3 cores
Then etching operation carries out expansive working twice to remove some small noises with 5*3 cores, then carry out an etching operation;
7) region labeling is carried out using the foreground image after connected domain calibration algorithm pair the 6) step Morphological scale-space, utilizes rectangle frame
Lock the moving target demarcated;
Step 2) using mixed Gaussian background modeling method extraction background image specific method be,
Using the gauss hybrid models of the K a certain pixel X of single gaussian probability model construction, see shown in formula (3);
Wherein, p (Xt) it is that t moment pixel value X occurstProbability, wi,tIt indicates the weights of i-th of Gauss model of t moment, and weighs
It is worth and indicates Gauss model sum for 1, K, takes 3-5, η (Xt,μi,t,∑i,t) indicate i-th of Gauss model of t moment, μi,tFor
Mean value, ∑i,tFor covariance matrix, n representation dimensions are shown in formula (4);
Mixture Gaussian background model matching is as follows with renewal process:
Model Matching be video image current frame pixel value X and existing K Gauss model are subjected to matching comparison, if i-th
Gauss model meets formula (5), then it represents that current frame pixel value is matching, otherwise mismatches;
|Xt-μi,t-1| 2.5 σ of <i,t-1 (5)
If matching is unsuccessful, the mean value of video present frame is used, and set a larger variance yields, establish new Gaussian Profile
Model;
The update of model is carried out according to formula (6) according to matching result;
Wherein, α indicates that video present frame is embedded into the rate of background model, referred to as learning rate, if Model Matching, Mi,t=
1, it is otherwise 0, μ and σ2It remains unchanged;
Due to ∑i,tThe smaller and big gaussian probability distributed model of weights is more likely used for approximate representation background pixel distributed mode
Type, for this purpose, to video per frame image in the sequence successively decreased according to the size of w/ σ values of pixel value to K gaussian probability distributed mode
Type sorts, and regard the distribution of preceding B gaussian probability as background, constitutes background image BI, see formula (7);
Wherein, τ is the threshold value of background model setting, τ value ranges [0.7,0.8].
2. the moving target detecting method of adaptive complex scene according to claim 1, it is characterised in that:Step 1)
Illumination compensation carries out as follows,
Assuming that I (t) indicates that inputted video image frame, δ indicate that two interframe allow the maximum global illumination occurred variation;It calculates first
The average pixel value of each frame sequence image of videoThen illumination compensation is carried out using following rule:
In formula, sgn () indicates sign function,Indicate the image after compensation.
3. the moving target detecting method of adaptive complex scene according to claim 1, it is characterised in that:Step 4) is most
Excellent segmentation threshold acquisition methods are,
If a width size is the image I (x, y) of M*N, I (x, y) indicates the grey scale pixel value of image coordinate point (x, y), and gray scale
Value value range is 0- (L-1), and the filtered absolute difference image of step 3) is DF (x, y), niIndicate absolute difference image
Gray value is the number of pixels of i, then number of pixels total amount is:piIndicate the probability of grey scale pixel value i, then:
pi=ni/ N, i=0,1,2,3 ..., L-1;
Then use segmentation candidates threshold value T that the pixel value in image is divided into two class of C0 and C1 by tonal gradation, C0 indicates target
Object, C1 indicate background object, i.e. C0={ 0,1 ..., T }, C1={ T+1, T+2 ..., L-1 }, then pixel corresponding to C0 and C1
Gray value probability distribution is respectively:
In formula,L is the number of gray level;So, the entropy of C0 and C1 is expressed from the next respectively;
On the basis of gained image C0 entropys and C1 entropys, then the sum of posterior entropy H is indicated as follows:
H=H0+H1;
So, compare to obtain tonal gradation corresponding to the maximum value of entropy-discriminate function, that is, indicate based on the optimal of maximum entropy algorithm
Segmentation threshold THR, is shown below,
Binary conversion treatment is carried out to filtered absolute difference image DF (x, y) using the optimum segmentation threshold value THR of acquisition, is obtained
Foreground image FI (x, y) in video, is shown below,
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510645189.4A CN105261037B (en) | 2015-10-08 | 2015-10-08 | A kind of moving target detecting method of adaptive complex scene |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510645189.4A CN105261037B (en) | 2015-10-08 | 2015-10-08 | A kind of moving target detecting method of adaptive complex scene |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105261037A CN105261037A (en) | 2016-01-20 |
CN105261037B true CN105261037B (en) | 2018-11-02 |
Family
ID=55100708
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510645189.4A Expired - Fee Related CN105261037B (en) | 2015-10-08 | 2015-10-08 | A kind of moving target detecting method of adaptive complex scene |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105261037B (en) |
Families Citing this family (25)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106205217B (en) * | 2016-06-24 | 2018-07-13 | 华中科技大学 | Unmanned plane automatic testing method based on machine vision and unmanned plane method of control |
CN106157303A (en) * | 2016-06-24 | 2016-11-23 | 浙江工商大学 | A kind of method based on machine vision to Surface testing |
CN106651923A (en) * | 2016-12-13 | 2017-05-10 | 中山大学 | Method and system for video image target detection and segmentation |
CN107563272B (en) * | 2017-06-14 | 2023-06-20 | 南京理工大学 | Target matching method in non-overlapping vision monitoring system |
CN107507221A (en) * | 2017-07-28 | 2017-12-22 | 天津大学 | With reference to frame difference method and the moving object detection and tracking method of mixed Gauss model |
CN107909608A (en) * | 2017-10-30 | 2018-04-13 | 北京航天福道高技术股份有限公司 | The moving target localization method and device suppressed based on mutual information and local spectrum |
CN108376406A (en) * | 2018-01-09 | 2018-08-07 | 公安部上海消防研究所 | A kind of Dynamic Recurrent modeling and fusion tracking method for channel blockage differentiation |
CN108550163A (en) * | 2018-04-19 | 2018-09-18 | 湖南理工学院 | Moving target detecting method in a kind of complex background scene |
CN108960253A (en) * | 2018-06-27 | 2018-12-07 | 魏巧萍 | A kind of object detection system |
CN109166137A (en) * | 2018-08-01 | 2019-01-08 | 上海电力学院 | For shake Moving Object in Video Sequences detection algorithm |
CN109145805B (en) * | 2018-08-15 | 2021-05-14 | 深圳市豪恩汽车电子装备股份有限公司 | Moving target detection method and system under vehicle-mounted environment |
CN109345472B (en) * | 2018-09-11 | 2021-07-06 | 重庆大学 | Infrared moving small target detection method for complex scene |
CN109784164B (en) * | 2018-12-12 | 2020-11-06 | 北京达佳互联信息技术有限公司 | Foreground identification method and device, electronic equipment and storage medium |
CN109727266A (en) * | 2019-01-08 | 2019-05-07 | 青岛一舍科技有限公司 | A method of the target person photo based on the pure view background of video acquisition |
CN109978917A (en) * | 2019-03-12 | 2019-07-05 | 黑河学院 | A kind of Dynamic Object Monitoring System monitoring device and its monitoring method |
CN110348305B (en) * | 2019-06-06 | 2021-06-25 | 西北大学 | Moving object extraction method based on monitoring video |
CN110472569A (en) * | 2019-08-14 | 2019-11-19 | 旭辉卓越健康信息科技有限公司 | A kind of method for parallel processing of personnel detection and identification based on video flowing |
CN111723634B (en) * | 2019-12-17 | 2024-04-16 | 中国科学院上海微***与信息技术研究所 | Image detection method and device, electronic equipment and storage medium |
CN111311644B (en) * | 2020-01-15 | 2021-03-30 | 电子科技大学 | Moving target detection method based on video SAR |
CN111583279A (en) * | 2020-05-12 | 2020-08-25 | 重庆理工大学 | Super-pixel image segmentation method based on PCBA |
CN111652910B (en) * | 2020-05-22 | 2023-04-11 | 重庆理工大学 | Target tracking algorithm based on object space relationship |
CN112446889A (en) * | 2020-07-01 | 2021-03-05 | 龚循安 | Medical video reading method based on ultrasound |
CN113052940B (en) * | 2021-03-14 | 2024-03-15 | 西北工业大学 | Space-time correlation map real-time construction method based on sonar |
CN114463389B (en) * | 2022-04-14 | 2022-07-22 | 广州联客信息科技有限公司 | Moving target detection method and detection system |
CN116182871B (en) * | 2023-04-26 | 2023-07-07 | 河海大学 | Sea cable detection robot attitude estimation method based on second-order hybrid filtering |
-
2015
- 2015-10-08 CN CN201510645189.4A patent/CN105261037B/en not_active Expired - Fee Related
Non-Patent Citations (5)
Title |
---|
一种新的最大熵阈值图像分割方法;陈鹏;《计算机科学》;20111130;第114-116页 * |
一种混合高斯背景模型下的像素分类运动目标检测方法;高凯亮等;《南京大学学报(自然科学)》;20110331;第47卷(第2期);第195-200页 * |
一种适应光照突变的运动目标检测方法;杨霖;《科技创新与应用》;20150718(第20期);第56-57页 * |
基于特征融合的粒子滤波目标跟踪新方法;闫河等;《光电子·激光》;20141031;第25卷(第10期);第1990-1999页 * |
运动目标检测的研究与DSP实现;孟介成;《中国优秀硕士学位论文全文数据库信息科技辑(月刊)》;20130515(第05期);第1-34页 * |
Also Published As
Publication number | Publication date |
---|---|
CN105261037A (en) | 2016-01-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105261037B (en) | A kind of moving target detecting method of adaptive complex scene | |
CN107204006B (en) | Static target detection method based on double background difference | |
CN106874949B (en) | Movement imaging platform moving target detecting method and system based on infrared image | |
CN107644429B (en) | Video segmentation method based on strong target constraint video saliency | |
CN102307274B (en) | Motion detection method based on edge detection and frame difference | |
CN103971386B (en) | A kind of foreground detection method under dynamic background scene | |
US8280165B2 (en) | System and method for segmenting foreground and background in a video | |
CN103258332B (en) | A kind of detection method of the moving target of resisting illumination variation | |
CN101299268B (en) | Semantic object dividing method suitable for low depth image | |
CN106846359A (en) | Moving target method for quick based on video sequence | |
CN107016691A (en) | Moving target detecting method based on super-pixel feature | |
Vosters et al. | Background subtraction under sudden illumination changes | |
CN107895379A (en) | The innovatory algorithm of foreground extraction in a kind of video monitoring | |
CN110751635B (en) | Oral cavity detection method based on interframe difference and HSV color space | |
CN106204594A (en) | A kind of direction detection method of dispersivity moving object based on video image | |
CN102096931A (en) | Moving target real-time detection method based on layering background modeling | |
CN106296744A (en) | A kind of combining adaptive model and the moving target detecting method of many shading attributes | |
CN102609945A (en) | Automatic registration method of visible light and thermal infrared image sequences | |
Lian et al. | A novel method on moving-objects detection based on background subtraction and three frames differencing | |
CN109166137A (en) | For shake Moving Object in Video Sequences detection algorithm | |
CN110807738A (en) | Fuzzy image non-blind restoration method based on edge image block sharpening | |
Liu et al. | Scene background estimation based on temporal median filter with Gaussian filtering | |
Zhang et al. | Dehazing with improved heterogeneous atmosphere light estimation and a nonlinear color attenuation prior model | |
CN110363197A (en) | Based on the video area-of-interest exacting method for improving visual background extraction model | |
CN109102520A (en) | The moving target detecting method combined based on fuzzy means clustering with Kalman filter tracking |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20181102 Termination date: 20191008 |
|
CF01 | Termination of patent right due to non-payment of annual fee |