CN102831617A - Method and system for detecting and tracking moving object - Google Patents
Method and system for detecting and tracking moving object Download PDFInfo
- Publication number
- CN102831617A CN102831617A CN2012102469467A CN201210246946A CN102831617A CN 102831617 A CN102831617 A CN 102831617A CN 2012102469467 A CN2012102469467 A CN 2012102469467A CN 201210246946 A CN201210246946 A CN 201210246946A CN 102831617 A CN102831617 A CN 102831617A
- Authority
- CN
- China
- Prior art keywords
- image
- zone
- tracked
- template
- target
- 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
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a moving target detecting and tracking method based on a DSP (digital signal processor), and a system for realizing the method. The method comprises the following steps of: firstly, establishing a background model by using a method that an interframe difference is combined with a background difference aiming at a Y-component of an image; subsequently detecting a foreground moving target; establishing an initial target template; carrying out pyramid downsampling for two times on the target template and a video image to be tracked respectively so as to reduce resolution of the target template and the video image to be tracked. A particle group optimization algorithm is adopted on a top layer of a pyramid to position a tracked target; a diamond search method is adopted on a middle layer and a bottom layer of the pyramid to position the tracked target; and the target template is updated constantly along with change of the moving target during a process that the target is tracked, thereby realizing real-time continuous tracking to the target.
Description
Technical field
The invention belongs to Flame Image Process and computer vision field, relate to the DSP technology, particularly the method and system of the moving object detection and tracking of gold tower multiresolution algorithm, particle swarm optimization algorithm and diamond search algorithm.
Background technology
Motion target detection and tracking technique are the important branch of computer vision, are the hot issues of current intelligent image processing and Video processing, and wide application prospect and long-range economic worth are arranged.At military aspect, the moving object detection and tracking technology can be used for motion target detection and tracking in aerial or the ground surveillance scope; Aspect social life, the moving object detection and tracking technology can be used for the real-time monitoring of various units, in time finds dangerous situation, the danger that prediction possibly occur; Aspect medical science, motion target detection and tracking technique can make the doctor better diagnose the state of an illness, remove less patient suffering.
Detection for Moving Target is meant through the computing between image sequence, finds out moving target region in the image, for the coupling and the tracking of moving target afterwards prepared.Moving target detecting method commonly used mainly contains: background subtraction point-score, frame-to-frame differences point-score and optical flow method.But the background subtraction point-score can't adapt to the variation of light in the actual environment; The cavity is often arranged in the moving region that splits under the frame-to-frame differences point-score; The calculated amount of optical flow method is huge, and complex algorithm can't satisfy the requirement of real-time follow-up.
The motion target tracking technology is meant the moving target that analyzing and testing is obtained, and the same moving target in the different frame is associated, and obtains its movement locus.Motion target tracking method commonly used has: based on the tracking of template, the tracking based on characteristic, the tracking in based target zone and the tracking of based target profile.But the following calculation amount that is based on template is very big; Tracking based on characteristic is difficult to confirm clarification of objective; The tracking in based target zone is followed easily when target is blocked or is out of shape and is lost target; The tracking of based target profile is difficult to the fast or big target of deformation of real-time follow-up speed.
Particle swarm optimization algorithm is that doctor Kennedy and doctor Eberhart pass through the research to the flock of birds foraging behavior; A kind of optimization method of finding the solution challenge that proposes; When finding the solution the optimum solution of problem; Each particle upgrades speed
and position
of oneself through current self the optimum solution (global extremum
) of optimum solution (individual extreme value
) and whole population; And then through the corresponding adaptive value of fitness evaluation function calculation; And then relatively adaptive value is selected total optimization and is separated; Upgrade follow-on speed and position according to following formula, thereby continuous iteration searches out the optimum solution of system.
Wherein,
is inertia weight;
is particle's velocity;
is the study factor; Its value is
;
and
is (0; 1) random number between;
represents individual extreme value, and
represents global extremum.Particle in the space constantly the experience renewal particle rapidity of the individual extreme value of study and global extremum and position up to searching out optimum solution.
Diamond search (ds) is meant the bigoted thought in the center that makes full use of motion vector, finds the optimum matching zone in object module and zone to be searched through near the point the search window center position.When using this method search optimum matching zone, select little search pattern possibly be absorbed in local optimum, and select big search pattern possibly can't find the optimum matching zone, so this method adopts the diamond template of two kinds of shape sizes to carry out match search.Its search procedure is following: at first the center from search window begins, and uses the bitellos template to search for repeatedly.If optimal match point appears at the edge of module, be the center with this point so, continue to search for the bitellos template; If optimal match point appears at the center of template, use the melee template so instead and search for four points on every side, find out optimal match point, finish search at last.In search procedure,, so just no longer consider these and put if this step needs the point of search in a last step, to search for.If some point has exceeded the scope of search window in the template, so also no longer consider these points.Though the number of times of the unqualified search of diamond search (ds), because the center-biased of motion vector can search the optimum matching zone very soon.
Need to propose more effectively moving object detection and tracking method at present.
Summary of the invention
For overcoming the deficiency of above-mentioned prior art problem, the present invention proposes a kind of real-time based on motion target detection and the tracking of DSP and the image processing system of realizing this method, thus the real-time detection and the tracking of realization moving target.
Moving object detection and tracking system based on DSP provided by the invention comprises video acquisition module, video processing module and display module; Said video acquisition module is through one road CCD camera collection vision signal; The analog video signal of gathering is transferred to SEED VPM642 video processing module; In VPM642, convert analog video signal the vision signal of BT.656 form to, and this signal is transferred to the video interface of DSP through high-performance Video Decoder TVP5150; The video interface of DSP combines the EDMA passage vision signal to be sent in the buffer area of SDRAM; After handling data through DM642; Become simulating signal to send the display of display module to data-switching by video encoder SAA7121H, through display display foreground moving target.
The present invention also provides a kind of improved background modeling method to Y component in the image that collects, and comprises the steps:
Step 1: at first import a two field picture, whether judge frame number, when frame number Y component to input picture greater than 3 time carries out continuous three-frame difference, and differentiated image is carried out binaryzation greater than 3; Three two field pictures of supposing input are respectively
; Its difference image is designated as
respectively; Wherein
; Its three-frame difference image is designated as
, and & represents logic and operation here;
Step 2:
carried out binary conversion treatment, generate binary image
;
Step 3: the module with binary image
is divided into
is designated as
;
Step 4: each piece
is carried out the morphology connected region detect and filtering, remove noise; If 0 number surpasses 85% of whole pixel number in should the zone, then the background of this piece image is stable, carry out step 5; Otherwise, carry out step 6;
Step 5: the background image model that generates this piece image;
Step 6: whether the quantity of judging the background module that generates is greater than 90% of integral module quantity; If greater than 90%; Then background model initializing finishes; Generation background iconic model
, otherwise, carry out step 1.
Moving object detection and tracking method based on DSP provided by the invention may further comprise the steps:
A, detect the foreground moving target through the method that inter-frame difference and background subtraction branch combine to the Y component of image;
B, set up To Template through the foreground moving target and confirm next frame zone to be tracked simultaneously;
C, the pyramid sampling is carried out in To Template and zone to be tracked;
D, through particle swarm optimization algorithm and diamond search algorithm keeps track moving target;
E, continual renovation To Template are realized the real-time follow-up of moving target.
Foregoing moving object detection and tracking method, preferred scheme is that said steps A concrete steps are following:
A1, set up background model through the method for inter-frame difference and the modeling of background difference to the Y component of image;
A2, the Y component of current frame image and the Y component of background image are subtracted each other, obtain the error image of image Y component;
A3, the error image of the Y component that obtained by A2 is carried out binary conversion treatment;
The binary image of A4, Y component that A3 is obtained carries out connected region and detects and filtering, removes noise;
The binary image of A5, Y component that A4 is obtained is sought out the foreground moving target respectively to x axle and y axial projection according to preset threshold.
Foregoing moving object detection and tracking method; Preferred scheme is; Said step B concrete steps are following: with in the Y component of current frame image with the centre of form of moving target as the center; The zone of choosing
is as To Template; The centre of form of the moving target that a two field picture records more than in the Y of next frame image component is the center, and the zone of choosing
is as zone to be tracked.
Foregoing moving object detection and tracking method, preferred scheme is that said step C concrete steps are following:
C1, sampling processing is fallen in To Template, obtain the middle layer pyramid diagram picture of To Template; Sampling processing falls in the middle layer pyramid to To Template then, obtains the top layer pyramid of To Template;
C2, treat tracing area and fall sampling processing, obtain the middle layer pyramid in zone to be tracked, sampling processing falls in the middle layer pyramid of treating tracing area then, obtains the top layer pyramid in zone to be tracked.
Foregoing moving object detection and tracking method, preferred scheme is that said step D concrete steps are following:
D1, the top layer pyramid in To Template and zone to be tracked is carried out matching operation; In the process of search, combine the PSO algorithm; Obtain the position of moving target in the regional top layer pyramid to be tracked through interparticle interaction; The relevant matches function that uses among the present invention is a minimum average B configuration absolute difference function (MAD), uses this function can reduce operand effectively;
D2, according to the matching area location of pixels that D1 obtains, in the image of the middle layer in zone to be tracked, carry out search matched, the zone of being searched is near the position of D1 Matching Location, the searching method that uses in the process of search is a diamond search (ds);
D3, according to the position that D2 confirms, in the original layers image in zone to be tracked, carry out search matched, the method for using in the process of search is a diamond search (ds), through this step, searches the position of moving target in this frame;
D4, according to the position that D3 confirms, go out the target location in this two field picture marked, mark the hunting zone of next frame simultaneously.
Foregoing moving object detection and tracking method, preferred scheme is that said step e concrete steps are following:
E1, according to the position that D4 confirms, upgrade the scope in zone to be tracked in To Template and the next frame image;
Whether E2, judgement tracking finish, if do not finish, then get back to the tracking that B1 carries out next frame.
Moving object detection and tracking method provided by the invention, this method comprises: gather sequence of video images, the Y component of image sequence through the method for inter-frame difference with the inter-frame difference combination, is set up the background image under the Y component; The Y component of current frame image and the Y component of background image are carried out the background difference, obtain the error image of current frame image Y component; Then the image Y component that obtains is carried out binaryzation, then it is carried out that connected region detects and filtering, remove noise, then with image respectively to x axle and y axial projection, seek out the foreground moving target according to preset threshold.
Moving object detection and tracking method provided by the invention; This method comprises: with the centre of form of the foreground moving target that searches out as the center; The zone of choosing
is as To Template; The centre of form of the moving target that a two field picture records more than in the Y of next frame image component is the center, and the zone of choosing
is as zone to be tracked.
Then To Template and zone to be tracked are carried out the pyramid sampling, concrete steps are following: sampling processing is fallen in To Template, obtain the middle layer pyramid diagram picture of To Template; Sampling processing falls in the middle layer pyramid to To Template then, obtains the top layer pyramid of To Template; Treat tracing area and fall sampling processing, obtain the middle layer pyramid in zone to be tracked, sampling processing falls in the middle layer pyramid of treating tracing area then, obtains the top layer pyramid in zone to be tracked.
Top layer pyramid to To Template and zone to be tracked carries out matching operation, in the process of search, combines the PSO algorithm, obtains the position of moving target in the regional top layer pyramid to be tracked through interparticle interaction.Utilization PSO algorithm can the optimization searching process, in zone to be tracked, effectively searches out the optimum matching zone of To Template.
Wherein,
is inertia weight;
is particle's velocity;
is the study factor; Its value is
;
and
is (0; 1) random number between;
represents individual extreme value, and
represents global extremum.Particle is constantly learnt the experience of individual extreme value and global extremum and is upgraded particle rapidity and position in the space.Know and search out optimum solution.In top layer gold tower coupling; The region of search of particle is exactly zone to be searched; The particle's velocity that in the process that particle was searched, defines to him is the displacement and the direction of particle; The expression with
; The displacement size and Orientation of
expression row wherein, the displacement size and Orientation of
expression row.The fitness evaluation function that uses among the present invention is a minimum average B configuration absolute difference function (MAD); Upgrade individual extreme value and colony's extreme value according to the fitness evaluation function, can in the top layer pyramid in zone to be tracked, search out moving target fast and effectively through this method.
According to the matching area location of pixels that obtains, in the image of the middle layer in zone to be tracked, carry out search matched, near the position of the regional top level of matched location of being searched, the searching method that uses in the process of search is a diamond search (ds); Then in the original layers image in zone to be tracked, carry out search matched, the method for using in the process of search is a diamond search (ds), through this step, searches the position of moving target in this frame; Go out the target location in this two field picture marked, mark the hunting zone of next frame simultaneously.
According to the target location that marks, upgrade the scope in zone to be tracked in To Template and the next frame image, then judge to follow the tracks of and whether finish,, then begin the tracking of next frame if do not finish.
The present invention adopts the tracking based on template, in order to reduce the complexity of calculating, has proposed to combine based on gold tower multiresolution the template matching method of particle swarm optimization algorithm and diamond search algorithm.The pyramid multiresolution just is meant through reducing the number of pixels downscaled images of image; Match search is carried out in To Template and target search zone after will dwindling on this basis; If after dwindling once; The calculated amount of search is still very big, then carries out second time image and dwindles, and the number of times that image dwindles is decided according to the size of To Template image and area image to be searched.If the image that will dwindle is arranged above and below according to the height of resolution, just form a pyramid structure, in this pyramid, the resolution of image that is positioned at the upper strata is low more, and the pixel count of image is just few more.
Method provided by the invention at first uses the method for inter-frame difference and the combination of background difference to set up background model to the Y component of image, detects the foreground moving target then; Then set up the initial target template, again To Template and video image to be tracked are carried out twice pyramid respectively and fall sampling, reduce the resolution of To Template and video image to be tracked through the foreground moving target.On top layer gold tower, adopt particle swarm optimization algorithm that tracking target is positioned, on middle layer and bottom pyramid, adopt the method for diamond search that tracking target is positioned.To Template is brought in constant renewal in the variation of moving target in the process of target following, realizes the real-time continuous of target is followed the tracks of.
Compared with prior art, technical advantage of the present invention also is embodied in:
1, the improved background modeling method to Y component in the input picture of the present invention's proposition just can be set up background model through 9 frames on the Y of image component, if the generation background image needed 30 frames could generate background model preferably after average was asked in employing.
2, the present invention is incorporated into gold tower multiresolution in motion target detection and the tracking, has effectively reduced the resolution of To Template and video image to be tracked.
3, the present invention is used in particle swarm optimization algorithm in the moving object detection and tracking system based on DSP first; Use this method can be in the top layer pyramid setting movement target fast and effectively; Its calculated amount is 90% of a common algorithm in practical implementation of the present invention, if bigger its calculated amount in zone to be tracked reduces obvious more.
4, the present invention is used in diamond search (ds) in the moving object detection and tracking system based on DSP; Use this method can in middle layer and bottom pyramid, locate tracking target fast; The overall calculation amount is 30% of a common templates matching method in practical implementation of the present invention, and accuracy can reach 92% under the illumination stable case.
Description of drawings
Fig. 1: image processing system figure.
Fig. 2: initialization background model.
Fig. 3: moving object detection and tracking algorithm flow chart.
Embodiment
Specify technical scheme of the present invention below in conjunction with embodiment and accompanying drawing, but protection domain is not by this restriction.
EmbodimentA kind of real-time based on motion target detection and the tracking of DSP and the image processing system of realizing this method, thus the real-time detection and the tracking of realization moving target.
Fig. 1 is the structural drawing based on the moving object detection and tracking system (image processing system) of DSP, and is as shown in Figure 1, and image processing system comprises video acquisition module, video processing module and three parts of display module.Pass through one road CCD camera collection vision signal among the present invention; The analog video signal of gathering is transferred to SEED VPM642 video processing module; In VPM642, convert analog video signal the vision signal of BT.656 form to, and this signal is transferred to the video interface of DSP through high-performance Video Decoder TVP5150.The video interface of DSP combines the EDMA passage vision signal to be sent in the buffer area of SDRAM; After handling data through DM642; Become simulating signal to send display to data-switching by video encoder SAA7121H, can see the effect of target following through display.
What the present invention proposed mainly comprises three parts based on the motion target detection of DSP and the workflow of tracker: DM642 initialization, system drive initialization, motion target detection and tracking.The DM642 initialization mainly comprises the initialization of chip internal memory interface, initializing peripheral equipment, interruption initialization etc.; The system drive initialization comprises Video Codec initialization, the initialization of DM642 video port, the initialization of EDMA passage etc.After accomplishing system initialization, DM642 no longer intervenes the input and output of vision signal, and system gets into the infinite loop stage, is mainly used at this stage D M642 and realizes motion target detection and tracking.
The present invention proposes a kind of improved background modeling method to Y component in the image that collects, and its algorithm steps is as shown in Figure 2:
Step 1: at first import a two field picture, whether judge frame number, when frame number Y component to input picture greater than 3 time carries out continuous three-frame difference, and differentiated image is carried out binaryzation greater than 3.Three two field pictures of supposing input are respectively
; Its difference image is designated as
respectively; Wherein
, its three-frame difference image is designated as
(& represents logic and operation) here.
Step 2:
carried out binary conversion treatment, generate binary image
.
Step 3: the module with binary image
is divided into
is designated as
.
Step 4: each piece
is carried out the morphology connected region detect and filtering, remove noise.If 0 number surpasses 85% of whole pixel number in should the zone, then the background of this piece image is stable changes 5 over to), otherwise, change 6 over to).
Step 6: the background image model that generates this piece image.
Step 7: whether the quantity of judging the background module that generates is greater than 90% of integral module quantity; If greater than 90%; Then background model initializing finishes; Generation background iconic model
, otherwise, change 1 over to).
The moving object detection and tracking method based on DSP that Fig. 3 proposes for the present invention, this scheme comprises the steps:
Steps A: the method that combines through inter-frame difference and background subtraction branch to the Y component of image detects the foreground moving target and (obtains gray level image through present frame
and background model
are carried out difference; Then the gray level image that obtains is generated binary image by certain threshold binarization; This image is carried out the morphology connected region detect and filtering, remove noise.I.e.
, wherein
is threshold value).
The binary image that obtains respectively to x axle and y axial projection, is searched out the foreground moving target through the grey level histogram method.
Summary is got up, and said steps A concrete steps are following:
A1, set up background model through the method for inter-frame difference and the modeling of background difference to the Y component of image;
A2, the Y component of current frame image and the Y component of background image are subtracted each other, obtain the error image of image Y component;
A3, the error image of the Y component that obtained by A2 is carried out binary conversion treatment;
The binary image of A4, Y component that A3 is obtained carries out connected region and detects and filtering, removes noise;
The binary image of A5, Y component that A4 is obtained is sought out the foreground moving target respectively to x axle and y axial projection according to preset threshold.
Step B: set up To Template through the foreground moving target and confirm that simultaneously next frame zone to be tracked is (with the centre of form of the foreground moving target that searches out as the center; The zone of choosing
is as To Template; The centre of form of the moving target that a two field picture records more than in the Y of next frame image component is the center, and the zone of choosing
is as zone to be tracked).
Summary is got up; Said step B concrete steps are following: with in the Y component of current frame image with moving target the centre of form as the center; The zone of choosing
is as To Template; The centre of form of the moving target that a two field picture records more than in the Y of next frame image component is the center, and the zone of choosing
is as zone to be tracked.
Step C: To Template and zone to be tracked are carried out the pyramid sampling, and concrete steps are following:
Sampling processing is fallen in To Template, obtain the middle layer pyramid diagram picture of To Template; Sampling processing falls in the middle layer pyramid to To Template then, obtains the top layer pyramid of To Template; Treat tracing area and fall sampling processing, obtain the middle layer pyramid in zone to be tracked, sampling processing falls in the middle layer pyramid of treating tracing area then, obtains the top layer pyramid in zone to be tracked.
Step D: (the top layer pyramid to To Template and zone to be tracked carries out matching operation through particle swarm optimization algorithm and diamond search algorithm keeps track moving target; In the process of search, combine the PSO algorithm, obtain the position of moving target in the regional top layer pyramid to be tracked through interparticle interaction.Utilization PSO algorithm can the optimization searching process, in zone to be tracked, effectively searches out the optimum matching zone of To Template).
Wherein,
is inertia weight;
is particle's velocity;
is the study factor; Its value is
;
and
is (0; 1) random number between;
represents individual extreme value, and
represents global extremum.Particle is constantly learnt the experience of individual extreme value and global extremum and is upgraded particle rapidity and position in the space.Know and search out optimum solution.In top layer gold tower coupling; The region of search of particle is exactly zone to be searched; The particle's velocity that in the process that particle was searched, defines to him is the displacement and the direction of particle; The expression with
; The displacement size and Orientation of
expression row wherein, the displacement size and Orientation of
expression row.The fitness evaluation function that uses among the present invention is a minimum average B configuration absolute difference function (MAD); Upgrade individual extreme value and colony's extreme value according to the fitness evaluation function, can in the top layer pyramid in zone to be tracked, search out moving target fast and effectively through this method.
According to the matching area location of pixels that obtains, in the image of the middle layer in zone to be tracked, carry out search matched, near the position of the regional top level of matched location of being searched, the searching method that uses in the process of search is a diamond search (ds); Then in the original layers image in zone to be tracked, carry out search matched, the method for using in the process of search is a diamond search (ds), through this step, searches the position of moving target in this frame; Go out the target location in this two field picture marked, mark the hunting zone of next frame simultaneously.
That summarizes says that said step D concrete steps are following:
D1, the top layer pyramid in To Template and zone to be tracked is carried out matching operation; In the process of search, combine the PSO algorithm; Obtain the position of moving target in the regional top layer pyramid to be tracked through interparticle interaction; The relevant matches function that uses among the present invention is a minimum average B configuration absolute difference function (MAD), uses this function can reduce operand effectively;
D2, according to the matching area location of pixels that D1 obtains, in the image of the middle layer in zone to be tracked, carry out search matched, the zone of being searched is near the position of D1 Matching Location, the searching method that uses in the process of search is a diamond search (ds);
D3, according to the position that D2 confirms, in the original layers image in zone to be tracked, carry out search matched, the method for using in the process of search is a diamond search (ds), through this step, searches the position of moving target in this frame;
D4, according to the position that D3 confirms, go out the target location in this two field picture marked, mark the hunting zone of next frame simultaneously.
Step: E: the real-time follow-up of bringing in constant renewal in To Template realization moving target is (according to the target location that marks; Upgrade the scope in zone to be tracked in To Template and the next frame image; Then judge to follow the tracks of whether finish,, then got back to for second step and carry out the tracking of next frame) if do not finish.
That summarizes says that said step e concrete steps are following:
E1, according to the position that D4 confirms, upgrade the scope in zone to be tracked in To Template and the next frame image;
Whether E2, judgement tracking finish, if do not finish, then get back to the tracking that B1 carries out next frame.
Claims (8)
1. based on the moving object detection and tracking system of DSP, it is characterized in that, comprise video acquisition module, video processing module and display module; Said video acquisition module is through one road CCD camera collection vision signal; The analog video signal of gathering is transferred to SEED VPM642 video processing module; In VPM642, convert analog video signal the vision signal of BT.656 form to, and this signal is transferred to the video interface of DSP through high-performance Video Decoder TVP5150; The video interface of DSP combines the EDMA passage vision signal to be sent in the buffer area of SDRAM; After handling data through DM642; Become simulating signal to send the display of display module to data-switching by video encoder SAA7121H, through display display foreground moving target.
2. an improved background modeling method to Y component in the image that collects is characterized in that, comprises the steps:
Step 1: at first import a two field picture, whether judge frame number, when frame number Y component to input picture greater than 3 time carries out continuous three-frame difference, and differentiated image is carried out binaryzation greater than 3; Three two field pictures of supposing input are respectively
; Its difference image is designated as
respectively; Wherein
; Its three-frame difference image is designated as
, and & represents logic and operation here;
Step 2:
carried out binary conversion treatment, generate binary image
;
Step 3: the module with binary image
is divided into
is designated as
;
Step 4: each piece
is carried out the morphology connected region detect and filtering, remove noise; If 0 number surpasses 85% of whole pixel number in should the zone, then the background of this piece image is stable, carry out step 5; Otherwise, carry out step 6;
Step 5: the background image model that generates this piece image;
Step 6: whether the quantity of judging the background module that generates is greater than 90% of integral module quantity; If greater than 90%; Then background model initializing finishes; Generation background iconic model
, otherwise, carry out step 1.
3. the moving object detection and tracking method based on DSP is characterized in that, may further comprise the steps:
A, detect the foreground moving target through the method that inter-frame difference and background subtraction branch combine to the Y component of image;
B, set up To Template through the foreground moving target and confirm next frame zone to be tracked simultaneously;
C, the pyramid sampling is carried out in To Template and zone to be tracked;
D, through particle swarm optimization algorithm and diamond search algorithm keeps track moving target;
E, continual renovation To Template are realized the real-time follow-up of moving target.
4. moving object detection and tracking method according to claim 3 is characterized in that, said steps A concrete steps are following:
A1, set up background model through the method for inter-frame difference and the modeling of background difference to the Y component of image;
A2, the Y component of current frame image and the Y component of background image are subtracted each other, obtain the error image of image Y component;
A3, the error image of the Y component that obtained by A2 is carried out binary conversion treatment;
The binary image of A4, Y component that A3 is obtained carries out connected region and detects and filtering, removes noise;
The binary image of A5, Y component that A4 is obtained is sought out the foreground moving target respectively to x axle and y axial projection according to preset threshold.
5. moving object detection and tracking method according to claim 3; It is characterized in that; Said step B concrete steps are following: with in the Y component of current frame image with moving target the centre of form as the center; The zone of choosing
is as To Template; The centre of form of the moving target that a two field picture records more than in the Y of next frame image component is the center, and the zone of choosing
is as zone to be tracked.
6. moving object detection and tracking method according to claim 3 is characterized in that, said step C concrete steps are following:
C1, sampling processing is fallen in To Template, obtain the middle layer pyramid diagram picture of To Template; Sampling processing falls in the middle layer pyramid to To Template then, obtains the top layer pyramid of To Template;
C2, treat tracing area and fall sampling processing, obtain the middle layer pyramid in zone to be tracked, sampling processing falls in the middle layer pyramid of treating tracing area then, obtains the top layer pyramid in zone to be tracked.
7. moving object detection and tracking method according to claim 3 is characterized in that, said step D concrete steps are following:
D1, the top layer pyramid in To Template and zone to be tracked is carried out matching operation; In the process of search, combine the PSO algorithm; Obtain the position of moving target in the regional top layer pyramid to be tracked through interparticle interaction; The relevant matches function that uses among the present invention is a minimum average B configuration absolute difference function (MAD), uses this function can reduce operand effectively;
D2, according to the matching area location of pixels that D1 obtains, in the image of the middle layer in zone to be tracked, carry out search matched, the zone of being searched is near the position of D1 Matching Location, the searching method that uses in the process of search is a diamond search (ds);
D3, according to the position that D2 confirms, in the original layers image in zone to be tracked, carry out search matched, the method for using in the process of search is a diamond search (ds), through this step, searches the position of moving target in this frame;
D4, according to the position that D3 confirms, go out the target location in this two field picture marked, mark the hunting zone of next frame simultaneously.
8. moving object detection and tracking method according to claim 3 is characterized in that, said step e concrete steps are following:
E1, according to the position that D4 confirms, upgrade the scope in zone to be tracked in To Template and the next frame image;
Whether E2, judgement tracking finish, if do not finish, then get back to the tracking that B1 carries out next frame.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012102469467A CN102831617A (en) | 2012-07-17 | 2012-07-17 | Method and system for detecting and tracking moving object |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012102469467A CN102831617A (en) | 2012-07-17 | 2012-07-17 | Method and system for detecting and tracking moving object |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102831617A true CN102831617A (en) | 2012-12-19 |
Family
ID=47334732
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2012102469467A Pending CN102831617A (en) | 2012-07-17 | 2012-07-17 | Method and system for detecting and tracking moving object |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102831617A (en) |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103577812A (en) * | 2013-11-22 | 2014-02-12 | 昆山视杰维光电科技有限公司 | Target recognition system and method |
| CN105096343A (en) * | 2015-08-14 | 2015-11-25 | 中国联合网络通信集团有限公司 | Method and device for tracking moving target |
| CN106296725A (en) * | 2015-06-12 | 2017-01-04 | 富泰华工业（深圳）有限公司 | Moving target detects and tracking and object detecting device in real time |
| CN106875420A (en) * | 2016-12-29 | 2017-06-20 | 北京理工雷科电子信息技术有限公司 | A kind of aerial moving-target sequence image emulation mode based on remote sensing platform |
| CN106997598A (en) * | 2017-01-06 | 2017-08-01 | 陕西科技大学 | The moving target detecting method merged based on RPCA with three-frame difference |
| CN107295296A (en) * | 2016-04-01 | 2017-10-24 | 中国科学院上海高等研究院 | A kind of selectively storage and restoration methods and system of monitor video |
| CN107356975A (en) * | 2017-07-12 | 2017-11-17 | 电子科技大学 | Underwater two-dimension space active electric-field localization method based on particle swarm optimization algorithm |
| CN107767396A (en) * | 2017-11-10 | 2018-03-06 | 广东工业大学 | Motion target tracking method, system, equipment and storage medium |
| CN108564787A (en) * | 2018-05-31 | 2018-09-21 | 北京理工大学 | Traffic observation procedure, system and equipment based on Floating Car method |
| CN108596129A (en) * | 2018-04-28 | 2018-09-28 | 武汉盛信鸿通科技有限公司 | A kind of vehicle based on intelligent video analysis technology gets over line detecting method |
| CN108900847A (en) * | 2018-06-11 | 2018-11-27 | 阿依瓦（北京）技术有限公司 | A kind of long-range prediction method based on Block- matching |
| CN109034136A (en) * | 2018-09-06 | 2018-12-18 | 湖北亿咖通科技有限公司 | Image processing method, device, picture pick-up device and storage medium |
| CN109146918A (en) * | 2018-06-11 | 2019-01-04 | 西安电子科技大学 | A kind of adaptive related objective localization method based on piecemeal |
| CN109658437A (en) * | 2018-11-01 | 2019-04-19 | 深圳神目信息技术有限公司 | A kind of method and device of quick detection moving object |
| CN109685827A (en) * | 2018-11-30 | 2019-04-26 | 南京理工大学 | A kind of object detecting and tracking method based on DSP |
| CN110133641A (en) * | 2019-04-19 | 2019-08-16 | 电子科技大学 | A kind of through-wall imaging radar target tracking method of dimension self-adaption |
| CN110610514A (en) * | 2018-06-15 | 2019-12-24 | 株式会社日立制作所 | Method, device and electronic equipment for realizing multi-target tracking |
| CN111294594A (en) * | 2020-02-26 | 2020-06-16 | 浙江大华技术股份有限公司 | Security inspection method, device, system and storage medium |
| CN111612819A (en) * | 2020-05-15 | 2020-09-01 | 安徽工程大学 | Moving target tracking method and system based on image pyramid absolute error sum |
| CN112352137A (en) * | 2018-05-25 | 2021-02-09 | 艾美创有限公司 | Motion encoder |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060291695A1 (en) * | 2005-06-24 | 2006-12-28 | Objectvideo, Inc. | Target detection and tracking from overhead video streams |
| CN101355692A (en) * | 2008-07-30 | 2009-01-28 | 浙江大学 | Intelligent monitoring apparatus for real time tracking motion target area |
| CN101540892A (en) * | 2009-04-23 | 2009-09-23 | 上海中安电子信息科技有限公司 | Method for people counting in doorway on DSP video gathering device |
-
2012
- 2012-07-17 CN CN2012102469467A patent/CN102831617A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060291695A1 (en) * | 2005-06-24 | 2006-12-28 | Objectvideo, Inc. | Target detection and tracking from overhead video streams |
| CN101355692A (en) * | 2008-07-30 | 2009-01-28 | 浙江大学 | Intelligent monitoring apparatus for real time tracking motion target area |
| CN101540892A (en) * | 2009-04-23 | 2009-09-23 | 上海中安电子信息科技有限公司 | Method for people counting in doorway on DSP video gathering device |
Non-Patent Citations (2)
| Title |
|---|
| 刘静: "固定场景下的运动目标检测算法与实现", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 葛广英: "基于神经网络的智能交通信息管理系统研究", 《电视技术》 * |
Cited By (32)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103577812A (en) * | 2013-11-22 | 2014-02-12 | 昆山视杰维光电科技有限公司 | Target recognition system and method |
| CN106296725A (en) * | 2015-06-12 | 2017-01-04 | 富泰华工业（深圳）有限公司 | Moving target detects and tracking and object detecting device in real time |
| TWI670684B (en) * | 2015-06-12 | 2019-09-01 | 鴻海精密工業股份有限公司 | A method for detecting and tracing a moving target and a target detection device |
| CN105096343B (en) * | 2015-08-14 | 2018-10-30 | 中国联合网络通信集团有限公司 | A kind of method for tracking moving target and device |
| CN105096343A (en) * | 2015-08-14 | 2015-11-25 | 中国联合网络通信集团有限公司 | Method and device for tracking moving target |
| CN107295296A (en) * | 2016-04-01 | 2017-10-24 | 中国科学院上海高等研究院 | A kind of selectively storage and restoration methods and system of monitor video |
| CN107295296B (en) * | 2016-04-01 | 2020-04-14 | 中国科学院上海高等研究院 | Method and system for selectively storing and recovering monitoring video |
| CN106875420A (en) * | 2016-12-29 | 2017-06-20 | 北京理工雷科电子信息技术有限公司 | A kind of aerial moving-target sequence image emulation mode based on remote sensing platform |
| CN106875420B (en) * | 2016-12-29 | 2019-08-23 | 北京理工雷科电子信息技术有限公司 | A kind of aerial moving-target sequence image emulation mode based on remote sensing platform |
| CN106997598A (en) * | 2017-01-06 | 2017-08-01 | 陕西科技大学 | The moving target detecting method merged based on RPCA with three-frame difference |
| CN107356975A (en) * | 2017-07-12 | 2017-11-17 | 电子科技大学 | Underwater two-dimension space active electric-field localization method based on particle swarm optimization algorithm |
| CN107356975B (en) * | 2017-07-12 | 2019-08-13 | 电子科技大学 | Underwater two-dimension space active electric-field localization method based on particle swarm optimization algorithm |
| CN107767396A (en) * | 2017-11-10 | 2018-03-06 | 广东工业大学 | Motion target tracking method, system, equipment and storage medium |
| CN108596129A (en) * | 2018-04-28 | 2018-09-28 | 武汉盛信鸿通科技有限公司 | A kind of vehicle based on intelligent video analysis technology gets over line detecting method |
| CN108596129B (en) * | 2018-04-28 | 2022-05-06 | 武汉盛信鸿通科技有限公司 | Vehicle line-crossing detection method based on intelligent video analysis technology |
| CN112352137A (en) * | 2018-05-25 | 2021-02-09 | 艾美创有限公司 | Motion encoder |
| US11885650B2 (en) | 2018-05-25 | 2024-01-30 | Imetrum Ltd. | Motion encoder |
| CN112352137B (en) * | 2018-05-25 | 2023-03-14 | 艾美创有限公司 | Motion encoder |
| CN108564787A (en) * | 2018-05-31 | 2018-09-21 | 北京理工大学 | Traffic observation procedure, system and equipment based on Floating Car method |
| CN108900847A (en) * | 2018-06-11 | 2018-11-27 | 阿依瓦（北京）技术有限公司 | A kind of long-range prediction method based on Block- matching |
| CN109146918A (en) * | 2018-06-11 | 2019-01-04 | 西安电子科技大学 | A kind of adaptive related objective localization method based on piecemeal |
| CN110610514A (en) * | 2018-06-15 | 2019-12-24 | 株式会社日立制作所 | Method, device and electronic equipment for realizing multi-target tracking |
| CN110610514B (en) * | 2018-06-15 | 2023-09-19 | 株式会社日立制作所 | Method, device and electronic equipment for realizing multi-target tracking |
| CN109034136A (en) * | 2018-09-06 | 2018-12-18 | 湖北亿咖通科技有限公司 | Image processing method, device, picture pick-up device and storage medium |
| CN109658437A (en) * | 2018-11-01 | 2019-04-19 | 深圳神目信息技术有限公司 | A kind of method and device of quick detection moving object |
| CN109685827A (en) * | 2018-11-30 | 2019-04-26 | 南京理工大学 | A kind of object detecting and tracking method based on DSP |
| CN109685827B (en) * | 2018-11-30 | 2022-09-06 | 南京理工大学 | Target detection and tracking method based on DSP |
| CN110133641B (en) * | 2019-04-19 | 2023-04-25 | 电子科技大学 | Scale-adaptive through-wall imaging radar target tracking method |
| CN110133641A (en) * | 2019-04-19 | 2019-08-16 | 电子科技大学 | A kind of through-wall imaging radar target tracking method of dimension self-adaption |
| CN111294594A (en) * | 2020-02-26 | 2020-06-16 | 浙江大华技术股份有限公司 | Security inspection method, device, system and storage medium |
| CN111294594B (en) * | 2020-02-26 | 2022-06-03 | 浙江华视智检科技有限公司 | Security inspection method, device, system and storage medium |
| CN111612819A (en) * | 2020-05-15 | 2020-09-01 | 安徽工程大学 | Moving target tracking method and system based on image pyramid absolute error sum |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102831617A (en) | Method and system for detecting and tracking moving object | |
| CN102184552B (en) | Moving target detecting method based on differential fusion and image edge information | |
| CN101751677B (en) | Target continuous tracking method based on multi-camera | |
| CN103997624B (en) | Overlapping domains dual camera Target Tracking System and method | |
| CN109785363A (en) | A kind of unmanned plane video motion Small object real-time detection and tracking | |
| CN103325126B (en) | A kind of video target tracking method under dimensional variation and circumstance of occlusion | |
| CN101826228B (en) | Detection method of bus passenger moving objects based on background estimation | |
| CN108961312A (en) | High-performance visual object tracking and system for embedded vision system | |
| WO2022001961A1 (en) | Detection method, detection device and detection system for moving target thrown from height | |
| CN102184551A (en) | Automatic target tracking method and system by combining multi-characteristic matching and particle filtering | |
| CN101344965A (en) | Tracking system based on binocular camera shooting | |
| CN103871079A (en) | Vehicle tracking method based on machine learning and optical flow | |
| CN101127887A (en) | Intelligent vision monitoring method and device | |
| CN103426183B (en) | Moving body track method and device | |
| CN106296725A (en) | Moving target detects and tracking and object detecting device in real time | |
| CN102332167A (en) | Target detection method for vehicles and pedestrians in intelligent traffic monitoring | |
| CN103279791A (en) | Pedestrian counting method based on multiple features | |
| CN105046719B (en) | A kind of video frequency monitoring method and system | |
| Chen et al. | Using FTOC to track shuttlecock for the badminton robot | |
| Lakshmi et al. | Neuromorphic vision: From sensors to event‐based algorithms | |
| CN102892007A (en) | Method and system for facilitating color balance synchronization between a plurality of video cameras as well as method and system for obtaining object tracking between two or more video cameras | |
| CN113763427B (en) | Multi-target tracking method based on coarse-to-fine shielding processing | |
| Chebli et al. | Pedestrian detection based on background compensation with block-matching algorithm | |
| CN102855637A (en) | Covariance tracking method for fusing bilateral filtering in video images | |
| Kyrkou | C 3 Net: end-to-end deep learning for efficient real-time visual active camera control |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20121219 |