CN103455997B - A kind of abandon detection method and system - Google Patents
A kind of abandon detection method and system Download PDFInfo
- Publication number
- CN103455997B CN103455997B CN201210180275.9A CN201210180275A CN103455997B CN 103455997 B CN103455997 B CN 103455997B CN 201210180275 A CN201210180275 A CN 201210180275A CN 103455997 B CN103455997 B CN 103455997B
- Authority
- CN
- China
- Prior art keywords
- abandon
- moving target
- tracing object
- tracing
- suspicious
- 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
Abstract
The present invention is applicable to safety monitoring technical field, and a kind of abandon detection method and system are provided. This detection method is extracted moving target and is followed the trail of from video sequence, and testing process is divided into doubtful abandon stage, preliminary judgement abandon stage and confirms abandon stage three phases, can avoid false alarm, and accuracy of detection is improved. In as the scene of the complexity such as dynamic background, multiple mobile object, can detect accurately abandon, and be easy to realize and application.
Description
Technical field
The invention belongs to safety monitoring technical field, relate in particular to a kind of abandon detection method and system.
Background technology
The situation is tense for world today's anti-terrorism, for ensureing social safety, the detection of suspicious abandon become to machineThe indispensable contents of public situation safety-protection system such as field, subway, traffic critical point and exhibition center. Although at presentThe CCTV camera extending as human vision ubiquity in business application, but current monitoring systemAs unification, can only record a video for evidence obtaining afterwards, and analysis video data Realtime Alerts that can not be real-time. Therefore,Develop there is the automaticity of practical significance, intelligent video monitoring system becomes urgent and necessary day by day.This just requires to replace human eye with video camera, and general-purpose computers contributor, replace people, to completeMonitor or monitor task.
The method that the detection of abandon mainly utilizes image to process and analyze automatically detects quiet in reality sceneState object, extracts object information (color, track, profile), and in real time and accurately recognition object is left over thingThe generation of part and warning, make computer have certain understanding and analyze video capability, thereby to hazard event toolThere is the function of active monitoring, strick precaution and warning. Here said abandon refer to passive movement main body carry from fortuneMove staticly, then separate the static object that exceedes certain hour and there is no affiliated main body with moving person.It is one of major function of intelligent video monitoring system that abandon detects, and studies for abandon detection algorithmTime is not long, but problem in practical application is a lot, therefore, is well solved not yet so far,Particularly how to improve the antijamming capability of algorithm, reduce the complexity of algorithm realization and improve the essence detectingDegree.
Summary of the invention
First technical problem to be solved by this invention is to provide a kind of abandon detection method, is intended to carryThe antijamming capability of high algorithm, reduces the complexity of algorithm realization and improves the precision detecting.
So the second technical problem that the present invention solves is to provide a kind of abandon detection system, be intended to improveThe antijamming capability of algorithm, reduces the complexity of algorithm realization and improves the precision detecting.
The present invention is achieved in that a kind of abandon detection method, comprises the steps:
Steps A, extracts moving target and follow the trail of at camera in the video sequence of taking;
Described steps A is followed the trail of the wish tracing movement target of extracting by following step:
Steps A 1, by do not reach life cycle default second life cycle threshold value moving target be defined as canDoubt tracing object;
Steps A 2, by reach life cycle default second life cycle threshold value moving target be defined as trackingObject;
Steps A 3, for the wish tracing movement target of extracting in each frame of video sequence, first judgement is fixedIn the tracing object that justice is crossed, whether there is the tracing object matching with it, if the tracing object having definedThe tracing object that middle existence matches with it, is associated this moving target with the tracing object matching;
Steps A 4, if steps A 3 judged results are do not exist in the tracing object having defined and extractWant the tracing object that tracing movement target matches, the suspicious tracing object that further judgement had definedIn whether there is the suspicious tracing object matching with the wish tracing movement target of extracting, if exist, by thisMoving target is associated with the suspicious tracing object matching, if do not exist, this moving target is defined asNew suspicious tracing object;
Steps A 5, each tracing object that first traversal defined, if there is moving target associated with it,Upgrade the life cycle of this tracing object, and the property parameters of this tracing object is updated to the fortune being associatedThe property parameters of moving-target; For the tracing object that does not have the moving target being associated, trigger execution stepB；
Steps A 6, after execution step A5, then travels through the each suspicious tracing object defining, if existMoving target associated with it, upgrades the life cycle of this suspicious tracing object, and judges this suspicious chasing afterWhether reach the life cycle of track object default second life cycle threshold value, if reach by this suspicious trackingObject definition is tracing object, the property parameters of this suspicious tracing object is updated to the motion being associated simultaneouslyThe property parameters of target, deletes for the suspicious tracing object that does not have the moving target being associated;
Step B, for the moving target of following the trail of in steps A, if there is displacement to be within a predetermined period of time less thanDefault displacement threshold value, and reach its life cycle default first life cycle threshold value, area be greater than defaultThe moving target of area threshold, this moving target is judged as to static doubtful abandon, and records thisThe position of moving target and colouring information;
Step C, for the moving target of following the trail of in steps A, if having moving target current video disappear andThis moving target belongs to the doubtful abandon of step B judgement, by the face of this moving target of step B recordIn look information and current video, the colouring information at same position place mates, if matching degree reaches defaultDegree of joining threshold value, tentatively confirms that this moving target is abandon;
Step D, if the time that the preliminary abandon of confirming of step C stops in video sequence exceedes defaultTime of staying threshold value, reports to the police.
Further, described step C also comprises the steps:
Step C1, for the moving target of following the trail of in steps A, if there is moving target to disappear at current videoAnd this moving target does not belong to the doubtful abandon of step B judgement, abandons following the trail of this moving target; Or
Step C2, for the moving target of following the trail of in steps A, if there is moving target to disappear at current videoAnd this moving target belongs to the doubtful abandon of step B judgement, this moving target step B being recordedIn colouring information and current video, the colouring information at same position place mates, default if matching degree does not reachMatching degree threshold value, abandon following the trail of this moving target and delete this doubtful abandon simultaneously.
Further, described step D specifically comprises the steps:
Step D1, for the preliminary abandon of confirming of step C, by this abandon is remembered in step BIn the colouring information of record and current video, the colouring information at same position place mates, pre-if matching degree reachesIf matching degree threshold value, confirm that this abandon is still present in current video;
Step D2, what statistic procedure D1 confirmed is still present in abandon in current video in video sequenceWhether the time stopping exceedes default time of staying threshold value, reports to the police if exceed.
Further, described step D specifically comprises the steps:
Step D3, for the preliminary abandon of confirming of step C, by this abandon is remembered in step BIn the colouring information of record and current video, the colouring information at same position place mates, if matching degree does not reachDefault matching degree threshold value, abandons following the trail of this abandon.
Further, in the tracing object that in described steps A 3, judgement had defined, whether exist and extractThe tracing object that matches of wish tracing movement target, described steps A 4 in judgement defined suspiciousWhen suspicious tracing object that the wish tracing movement target whether existing in tracing object and extract matches, all logicalCrossing following steps judges:
Steps A 7, judges moving target and tracing object, or the external square of moving target and suspicious tracing objectWhether shape intersects;
Steps A 8, if steps A 7 judged results are for intersecting, further whether the crossing area of judgement reaches pre-If area change rate or the moving target of boundary rectangle of crossing area threshold, moving target and tracing objectWith the area change rate of the boundary rectangle of suspicious tracing object whether within the scope of default rate of change;
Steps A 9, if steps A 8 judged results are to intersect area to reach default crossing area threshold, externalThe area change rate of rectangle, within the scope of default rate of change, judges the phase of moving target and tracing objectJoin, or the matching of moving target and suspicious tracing object.
Further, described steps A is extracted moving target by following step:
Steps A 10, what the mode combining by three-frame difference and mixed Gauss model was taken at camera looksIn frequency sequence, extract moving target; And/or,
Steps A 11, carries out morphology processing to the wish tracing movement target image extracting, so that cannot extract completelyThe whole areas of disconnection UNICOM causing. Abandon detection system provided by the invention comprises:
Moving object detection unit, extracts moving target and follows the trail of for the video sequence of taking at camera;
Described moving object detection unit comprises:
Suspicious tracing object memory module, for store do not reach life cycle default second life cycle thresholdThe moving target of value;
Tracing object memory module, for store reach life cycle default second life cycle threshold value fortuneMoving-target;
The first relating module, for the wish tracing movement target that each frame of video sequence is extracted, first sentencesIn tracing object in disconnected tracing object memory module, whether there is the tracing object matching with it, if existThe tracing object matching with it, is associated this moving target with the tracing object matching;
The second relating module, for being described tracing object storage mould in described the first relating module judged resultThe tracing object that the wish tracing movement target not existing in piece and extract matches, further judgement is described canWhether doubt in tracing object memory module exists the suspicious tracking matching with the wish tracing movement target of extracting rightResemble, if exist, this moving target is associated with the suspicious tracing object matching, if do not exist,This moving target is stored to described suspicious tracing object memory module;
Tracing object is new module more, right for first traveling through each tracking of described tracing object memory moduleResemble, if there is moving target associated with it, the life cycle of upgrading this tracing object, and this is chased afterThe property parameters of track object is updated to the property parameters of the moving target being associated; Be associated for not existingThe tracing object of moving target, triggers described doubtful abandon judging unit and judges;
More new module of suspicious tracing object, after in described tracing object, more new module is upgraded, then timeGo through the each suspicious tracing object in described suspicious tracing object memory module, if there is fortune associated with itMoving-target, upgrades the life cycle of this suspicious tracing object, and judges the existence week of this suspicious tracing objectPhase whether reach default second life cycle threshold value, if reach, this suspicious tracing object is defined as to trackingObject is updated to the property parameters of this suspicious tracing object the property parameters of the moving target being associated simultaneously,Delete for the suspicious tracing object that does not have the moving target being associated;
Doubtful abandon judging unit, for the moving target for described moving object detection cell tracks,Judge whether that displacement is within a predetermined period of time less than default displacement threshold value, and reach its life cycle pre-If first life cycle threshold value, area be greater than the moving target of default area threshold, if having by this fortuneMoving-target is judged as static doubtful abandon, and records position and the colouring information of this moving target;
Abandon preliminary judgement unit, for the moving target for described moving object detection cell tracks,Judge whether that moving target belongs to described doubtful abandon judgement list in current video disappearance and this moving targetThe doubtful abandon of unit's judgement, this moving target described doubtful abandon judging unit being recorded if havingIn colouring information and current video, the colouring information at same position place mates, default if matching degree reachesMatching degree threshold value, tentatively confirms that this moving target is abandon;
Abandon determining unit, is looking for the abandon of the preliminary confirmation in described abandon preliminary judgement unitThe time stopping in frequency sequence exceedes default time of staying when threshold value, reports to the police.
Further, described abandon preliminary judgement unit is also in judged result being described moving target inspectionThe moving target that measurement unit is followed the trail of disappears at current video, and this moving target does not belong to described doubtful abandonWhen the doubtful abandon of judging unit judgement, control described moving object detection unit and abandon following the trail of this motion orderMark.
Further, described abandon preliminary judgement unit is also in judged result being described moving target inspectionThe moving target that measurement unit is followed the trail of disappears at current video, and this moving target belongs to described doubtful abandon and sentencesWhen the doubtful abandon of disconnected unit judges, by this moving target of described doubtful abandon judging unit recordIn colouring information and current video, the colouring information at same position place mates, default if matching degree does not reachMatching degree threshold value, control described moving object detection unit and abandon following the trail of this moving target, and simultaneously deleteThis doubtful abandon.
Further, described abandon determining unit comprises:
The first matching module, for by the preliminary described abandon preliminary judgement unit abandon of confirming by instituteThe colouring information of stating same position place in the colouring information of doubtful abandon judging unit record and current video entersRow coupling, if matching degree reaches default matching degree threshold value, confirms that this abandon is still present in current videoIn;
Alarm module, the abandonment that is still present in current video of confirming for adding up described the first matching moduleWhether the time that thing stops in video sequence exceedes default time of staying threshold value, reports to the police if exceed.
Further, when the result after described the first matching module coupling is that matching degree does not reach default couplingWhen degree threshold value, described the first matching module also abandons following the trail of this fortune for controlling described moving object detection unitMoving-target, and delete this doubtful abandon simultaneously.
Further, first described the first relating module and the second relating module all judge moving target and trackingObject, or whether moving target is crossing with the boundary rectangle of suspicious tracing object, when judged result is for intersecting,Whether the crossing area of further judgement reaches default crossing area threshold, moving target and tracing objectThe area change rate of the boundary rectangle of the area change rate of boundary rectangle or moving target and suspicious tracing object isNo within the scope of default rate of change; If judged result be intersect area reach default crossing area threshold,The area change rate of boundary rectangle, within the scope of default rate of change, judges moving target and tracing objectMatch, or the matching of moving target and suspicious tracing object.
Further, described moving object detection unit comprises:
Moving target extraction module, is taking the photograph for the mode combining by three-frame difference and mixed Gauss modelIn the video sequence that picture head is taken, extract moving target; And/or,
Moving target processing module, for the wish tracing movement target image extracting is carried out to morphology processing,So that extract the imperfect areas of disconnection UNICOM causing.
The present invention utilizes the characteristic of abandon, first abandon is detected and be divided into the doubtful abandon stage, preliminaryJudge the abandon stage and confirm abandon stage three phases, can avoid false alarm, accuracy of detection is obtainedImprove, in addition, whole testing process is taking time index and range index as basis for estimation, and algorithm is simple and easy to realityExisting, also can select only the moving target entering completely in scene to be carried out to trace detection simultaneously, and ignore from lookingFrequently the interference of the moving object of edge process, contributes to provide the antijamming capability of whole trace detection computing.
Brief description of the drawings
Fig. 1 is the realization flow figure of the abandon detection method that provides of the embodiment of the present invention;
Fig. 2 is the flow chart that moving target is followed the trail of that the embodiment of the present invention provides;
Fig. 3 is the structure principle chart of the abandon detection system that provides of the embodiment of the present invention;
Fig. 4 is the structure principle chart of abandon determining unit in Fig. 3;
Fig. 5 is the structure principle chart of moving object detection unit in Fig. 3.
Detailed description of the invention
In order to make object of the present invention, technical scheme and advantage clearer, below in conjunction with accompanying drawing and realityExecute example, the present invention is further elaborated. Only should be appreciated that specific embodiment described hereinOnly, in order to explain the present invention, be not intended to limit the present invention.
The present invention extracts after moving target tracking from video sequence, and testing process is divided into doubtful abandonStage, preliminary judgement abandon stage and confirmation abandon stage three phases, while finally confirming as abandonJust report to the police.
Fig. 1 shows the realization flow of the abandon detection method that the embodiment of the present invention provides, and details are as follows.
In steps A, in the video sequence of taking at camera, extract moving target and follow the trail of.
Before carrying out abandon detection, first need to carry out the detection of moving target, be divided into again moving targetThe process of extracting from background and tracing process two aspects.
In step B, for the moving target of following the trail of in steps A, if there is displacement within a predetermined period of timeBe less than default displacement threshold value, and reach its life cycle default first life cycle threshold value, area be greater thanThe moving target of default area threshold, is judged as this moving target static doubtful abandon, and noteRecord position and the colouring information of this moving target.
The object that abandon detects is to find actionless tracing object, and this object can chase after by inquiryThe historical position information realization of track object. In the embodiment of the present invention, be divided into three rank for the detection of abandonSection, this step is the first stage: doubtful abandon detection-phase. The moving target of following the trail of in steps A is openedBegin certain hour static, but do not entered or enter completely in video background model, be now called doubtful abandonmentIn the thing stage, the target that this stage is found is called doubtful abandon. So-called doubtful abandon refers to become something lostThe moving target of gurry, the doubtful abandon in this stage has completely static and there is no a characteristic of deformation.
In the time of tracing object information updating, for each tracing object, do as judged:
A) whether certain hour intrinsic displacement is less than default displacement threshold value recently:Wherein,For tracing object pXPosition in k frame video is with respect at k-TkPosition in frame videoThe displacement of putting, elapsed time is T veryk,For default displacement threshold value.
B) life cycle n whether reach certain threshold value >=
C) area SXWhether be greater than default area threshold >=
If tracing object meets above condition, and be not defined as doubtful abandon, can judge this trackingObject meets the abandon condition of detection first stage, i.e. doubtful abandon, record this doubtful abandon quietPosition and colouring information only time, for second and the abandon of phase III detect, specifically can adopt straightThe form of side figure records colouring information. When concrete enforcement, can be doubtful abandon a doubtful abandon is setList, is all stored in all doubtful abandons in this doubtful abandon list.
In step C, for the moving target of following the trail of in steps A, if there is moving target to disappear at current videoMistake and this moving target belong to the doubtful abandon of step B judgement, by this moving target of step B recordColouring information and current video in the colouring information at same position place mate, default if matching degree reachesMatching degree threshold value, tentatively confirm that this moving target is abandon.
In the time that tracing object disappears, have two kinds may: one is due to long-time static and learned into background,Now cause following the trail of unsuccessfully owing to losing foreground image; Another kind may be because tracing object is being looked reallyFrequently in, disappeared, such as being blocked or be moved out of scene by other objects. For these two kinds possibilities, disappear at objectMiss the season, search doubtful abandon list.
If this object exists in doubtful abandon list, by the histogram information of doubtful abandon with work as forward sightFrequently on frame, the colouring information at same position place mates, and it is passable that matching degree reaches default matching degree threshold valueDetermine that doubtful abandon object is learnt into background, particularly, when both parameters of colouring informationDifference in allowed limits, can be thought the matching degree threshold value that reaches default, and both match. At this moment shouldDoubtful abandon meets the second stage decision rule that abandon detects, and can preliminary judgement be abandon. InBe, doubtful abandon is deleted from doubtful abandon list, can add an abandon list. Matching degreeThink that lower than default matching degree threshold value doubtful abandon is not in this position, doubtful abandon is blockedOr remove, now doubtful abandon should be left out from doubtful abandon list, abandon following the trail of this moving target.
If this object does not exist in doubtful abandon list, illustrate that this object disappears in video, nowShould directly delete this object, abandon following the trail of this moving target.
In step D, if the time that the preliminary abandon of confirming of step C stops in video sequence exceed in advanceIf time of staying threshold value, report to the police.
In the time of frame of video information updating, for each second stage of judging through abandon, be present in abandonmentAbandon in thing list, because the time of staying not yet reaches abandon time of fire alarming threshold value, now should continueThis abandon is followed the trail of, until exceed abandon alarm threshold value its quiescent time. In this stage, specifically pass throughTime index and range index are as basis for estimation, and wherein, time index is used for calculating temporary transient stationary object and stopsThe time of staying, when this object time of staying accumulation reaches certain threshold value (the present invention is made as 6s), send something lostGurry alarm; Range index has or not motion for judging temporary transient static object, because this resting is known from experience quiltMixture Gaussian background model is slowly learned in background, and less skew can occur its barycenter, quiet in the present inventionDecision rule is only rule of thumb that 60 frame intrinsic displacements are less than 10 pixels.
But because this abandon is learned in background, this abandon not in foreground image, the present invention is logicalCross and also need first to judge whether abandon still exists in current video, and concrete, this step D comprises:
Rapid D1, for the preliminary abandon of confirming of step C, by by this abandon record in step BColouring information and current video in the colouring information at same position place mate, default if matching degree reachesMatching degree threshold value, confirm that this abandon is still present in current video;
Step D2, what statistic procedure D1 confirmed is still present in abandon in current video in video sequenceWhether the time stopping exceedes default time of staying threshold value, reports to the police if exceed.
Further, also can comprise step D3, for the preliminary abandon of confirming of step C, by by thisIn the colouring information that abandon records in step B and current video, the colouring information at same position place carries outJoin, if matching degree does not reach default matching degree threshold value, abandon following the trail of this abandon, and delete this simultaneouslyThe preliminary abandon of confirming.
Only to be moved into actionless object after video because abandon detects what be concerned about, and moving targetWhether changing inactive state into needs trace detection, looks for exactly in each two field picture of one section of image sequenceTo the residing position of interested target, only has target to be traced to just to know target in whole videoMotion state, could further analyze the behavior of knowing target, for detecting, abandon prepares, wherein, noWith moving target there is different parameters, for example people and automobile wide, just different at high proportion, " interested "Moving target can be by user's self-defining, for example define the wide, at high proportion of moving target, only to wide,Meeting at high proportion the Moving Objects imposing a condition just follows the trail of. What at present, target tracking field was comparatively ripe isCamshift algorithm, but this algorithm blocks or deformation in the situation that at moving target, or moving target andIn the close situation of background color, easily cause following the tracks of unsuccessfully. In the present invention, because detecting, abandon is concerned aboutBe to be only moved into actionless object after video, therefore adopt the tracking of judging based on rectangle intersection areaAlgorithm, this algorithm is realized comparatively simple, and the effect and the real-time that detect for abandon are better. This targetTrack algorithm is the classification Processing Algorithm to tracing object, is mainly divided into new object generation and frame of video and upgradesTwo parts. Therefore further, when the wish tracing movement target of extracting tracking for steps A, canCarry out in the following way, specifically referring to Fig. 2:
In the present invention, based on moving target X definition tracing objectThe existence that wherein n is objectPhase is to be created and to start till the time that present frame experiences is also the frame number that object exists from object;For the parameter current of object, the parameter with moving target in present frame is consistent. For avoiding due to noiseThe impact that the interim moving target causing with the factor such as background perturbation detects abandon, by right all trackingsResemble according to its life cycle n and threshold value TeRelation be divided into two classes: interim tracing object and tracing object, respectivelyBe stored in corresponding list. Interim tracing object list is defined asTracing objectList is defined asIn interim tracing object list, store all suspicious trackings rightResemble, and only the tracing object in tracing object list is carried out to abandon detection algorithm. Be embodied in followingIn steps A 1 and A2.
Steps A 1, by do not reach life cycle default second life cycle threshold value moving target be defined as canDoubt tracing object.
Steps A 2, by reach life cycle default second life cycle threshold value moving target be defined as trackingObject.
Target tracking algorithm of the present invention is the classification Processing Algorithm to tracing object, is mainly divided into new objectGenerate and two parts of frame of video renewal. Carry out new object generation for Part I, for fear of some fromThe interference of the moving object of video edge process, only processes the moving target that enters scene completely. RightMoving target in k frame, investigates itself and V according to the adaptation function of formula (3) definitiontraceAnd VtempIn instituteThere is tracing object whether to have matching relationship. If moving target X and certain tracing objectMatch,Record the associated γ of the two:If there is not the tracing object of mating with moving target X,Generate new tracing object based on this moving targetAnd this object is added to Vtemp. Corresponding underState steps A 3 and A4.
Steps A 3, for the wish tracing movement target of extracting in each frame of video sequence, first judgement is fixedIn the tracing object that justice is crossed, whether there is the tracing object matching with it, if the tracing object having definedThe tracing object that middle existence matches with it, is associated this moving target with the tracing object matching.
Steps A 4, if steps A 3 judged results are do not exist in the tracing object having defined and extractWant the tracing object that tracing movement target matches, the suspicious tracing object that further judgement had definedIn whether there is the suspicious tracing object matching with the wish tracing movement target of extracting, if exist, by thisMoving target is associated with the suspicious tracing object matching, if do not exist, this moving target is defined asSuspicious tracing object.
When Part II carries out frame of video renewal, first travel through VtraceIn each tracing object, if exist withThe moving target X being associated meet γ:Upgrade the information of this tracing objectOtherwise, think that this tracing object disappears, trigger abandon and detect event.Then travel through VtempIn each tracing object, if there is moving target X associated therewith, upgrade shouldThe information of tracing object, and further judge its life cycle, whether n+1 met threshold value (T hereine＝10)Requirement, to determine whether this object is moved to VtraceIn; Otherwise, think that this tracing object disappears, directly from VtempThis object of middle deletion. Corresponding to following steps A 5 and A6.
Steps A 5, each tracing object that first traversal defined, if there is moving target associated with it,Upgrade the life cycle of this tracing object, and the property parameters of this tracing object is updated to the fortune being associatedThe property parameters of moving-target; For the tracing object that does not have the moving target being associated, trigger execution stepB。
Steps A 6, after execution step A5, then travels through the each suspicious tracing object defining, if existMoving target associated with it, upgrades the life cycle of this suspicious tracing object, and judges this suspicious chasing afterWhether reach the life cycle of track object default second life cycle threshold value, if reach by this suspicious trackingObject definition is tracing object, the property parameters of this suspicious tracing object is updated to the motion being associated simultaneouslyThe property parameters of target, deletes for the suspicious tracing object that does not have the moving target being associated.
Further, the principle that above-mentioned steps A3 and steps A 4 are mated is as follows:
Steps A 7, judges moving target and tracing object, or the external square of moving target and suspicious tracing objectWhether shape intersects;
Steps A 8, if steps A 7 judged results are for intersecting, further whether the crossing area of judgement reaches pre-If area change rate or the moving target of boundary rectangle of crossing area threshold, moving target and tracing objectWith the area change rate of the boundary rectangle of suspicious tracing object whether within the scope of default rate of change;
Steps A 9, if steps A 8 judged results are to intersect area to reach default crossing area threshold, externalThe area change rate of rectangle, within the scope of default rate of change, judges the phase of moving target and tracing objectJoin, or the matching of moving target and suspicious tracing object.
Illustrate, for each moving target that the target detection stage extracts, we can think eachConnected region is all a moving target, and calculates the boundary rectangle of moving target, the center of rectangular areaPosition represents the position of this target in video scene. Now, for a certain motion order in k two field pictureMark A, can obtain the following parameter of its boundary rectangle:
Be respectively wide, high, center x coordinate, center y coordinate, the area of the boundary rectangle of A. Obviously havesA＝wA×hA. For two moving target A in k frame, l two field picture and B respectively, judge outside itConnecing the discriminant function whether rectangle intersect is
For crossing target, can further judge whether coupling according to the rate of change that intersects area and area
Track phase can adopt takes out the method that frame detects, because the time interval is very little, full between adjacent two framesThe moving target of foot adaptation function is same target.
As described above, what first abandon detection needed solution is the extraction of Moving Object in Video Sequences,Background modeling has adopted mixture Gaussian background model, and it extracts movement destination image by setting up background model,But traditional mixed Gauss model exists some problems: as light sudden change can produce larger to foreground imageImpact, the moving target close with background color not can completely detects etc. In order to address the above problem,In the video sequence that we take at camera by three-frame difference and the mixture Gaussian background model mode that combinesExtract moving target, can play to eliminate and disturb and the accurately effect of foreground image, make the motion order extractingIt is more accurate to mark, and wherein, three-frame difference method is the one of frame difference method, adopts continuous three two field pictures to carry out differenceComputing, first calculates respectively the difference result between front cross frame image and rear two two field pictures, then by two differenceResult binaryzation, finally by two result phases with, can obtain the profile of moving target, this algorithm nowThrough very ripe, repeat no more herein. Each in mixture Gaussian background model K Gauss model for pixelWeighting represents, this algorithm is also very ripe, repeats no more. In the present invention, use first respectively three-frame difference methodExtract moving target with mixture Gaussian background model, then extract result phase "or" by two, asThe moving target extracting eventually. Finally, the wish tracing movement target image extracting is carried out to morphology processing, withMake to extract the imperfect areas of disconnection UNICOM causing.
One of ordinary skill in the art will appreciate that the whole or portion realizing in the method that the various embodiments described above provideCan carry out the hardware that instruction is relevant by program step by step and complete, described program can be stored in a calculatingIn machine read/write memory medium, this storage medium can be ROM/RAM, disk, CD etc.
Fig. 3 shows the structural principle of the abandon detection system that the embodiment of the present invention provides, for the ease of retouchingState, only show the part relevant to the present embodiment.
With reference to Fig. 3, the abandon detection system that the embodiment of the present invention provides comprise moving object detection unit 1,Doubtful abandon judging unit 2, abandon preliminary judgement unit 3, abandon determining unit 4, wherein, fortuneIn the video sequence that moving-target detecting unit 1 is taken at camera, extract moving target and follow the trail of doubtful abandonmentThe moving target that thing judging unit 2 is followed the trail of for moving object detection unit 1, has judged whether in pre-timingBetween displacement in section be less than default displacement threshold value, and reach its life cycle default first life cycle thresholdValue, area are greater than the moving target of default area threshold, if having, this moving target are judged as staticDoubtful abandon, and record position and the colouring information of this moving target. Abandon preliminary judgement unit 3The moving target of following the trail of for moving object detection unit 1, has judged whether that moving target disappears at current videoMistake and this moving target belong to the doubtful abandon that doubtful abandon judging unit 2 judges, if having by doubtfulThe face at same position place in the colouring information of this moving target that abandon judging unit 2 records and current videoLook information is mated, if matching degree reaches default matching degree threshold value, tentatively confirms that this moving target isAbandon. Finally, abandon determining unit 4 is at the preliminary abandon of confirming in abandon preliminary judgement unit 3The time stopping in video sequence exceedes default time of staying when threshold value, reports to the police. Above-mentioned each unitThe principle that judges, mates as described above, repeats no more herein.
Further, above-mentioned abandon preliminary judgement unit 3 is also for being moving object detection in judged resultThe moving target that unit 1 is followed the trail of disappears at current video, and this moving target does not belong to doubtful abandon judgementWhen doubtful abandon that unit 2 judges, controlled motion object detection unit 1 abandons following the trail of this moving target.
Further, above-mentioned abandon preliminary judgement unit 3 is also for being moving object detection in judged resultThe moving target that unit 1 is followed the trail of disappears at current video, and this moving target belongs to doubtful abandon judgement listUnit 2 judgement doubtful abandon time, the color of this moving target that doubtful abandon judging unit 2 is recordedIn information and current video, the colouring information at same position place mates, if matching degree does not reach defaultDegree of joining threshold value, controlled motion object detection unit 1 abandons following the trail of this moving target, and deletes that this is doubtful simultaneouslyAbandon.
Further, as shown in Figure 4, described abandon determining unit 4 comprises: the first matching module 41With alarm module 42, the first matching modules 41 by preliminary abandon preliminary judgement unit 3 abandon of confirmingThe colouring information being recorded by doubtful abandon judging unit 2 and current video in the color at same position place believeBreath mates, if matching degree reaches default matching degree threshold value, confirms that this abandon is still present in currentIn video; Alarm module 42 is still present in current video for what add up that the first matching module 41 confirmsWhether the time that abandon stops in video sequence exceedes default time of staying threshold value, carries out if exceedReport to the police.
Further, when the result after the first matching module 41 couplings is that matching degree does not reach default matching degreeWhen threshold value, the first matching module 41 also abandons following the trail of this motion order for controlling described moving object detection unitMark.
Further, as shown in Figure 5, moving object detection unit 1 comprises: suspicious tracing object storage mouldPiece 11, tracing object memory module 12, the first relating module 13, the second relating module 14, tracing objectMore more new module 16 of new module 15, suspicious tracing object. Wherein, suspicious tracing object memory module 11 useIn storage do not reach life cycle default second life cycle threshold value moving target, tracing object storage mouldPiece 12 store reach life cycle default second life cycle threshold value moving target.
The first relating module 13 is to the wish tracing movement target of extracting in each frame of video sequence, and first judgement chases afterIn tracing object in track object memory module 12, whether there is the tracing object matching with it, if exist withThe tracing object that it matches, is associated this moving target with the tracing object matching. The second associationModule 14 is in tracing object memory module 12, do not exist and extract in the first relating module 13 judged resultsWant the tracing object that tracing movement target matches, further judge described suspicious tracing object memory moduleThe suspicious tracing object that the wish tracing movement target whether existing in 11 and extract matches, if exist, willThis moving target is associated with the suspicious tracing object matching, if do not exist, by this moving target storageTo described suspicious tracing object memory module.
First tracing object more new module 15 travels through the each tracing object in tracing object memory module 12,If there is moving target associated with it, upgrade the life cycle of this tracing object, and by right this trackingThe property parameters of elephant is updated to the property parameters of the moving target being associated; For there not being the motion being associatedThe tracing object of target, triggers described doubtful abandon judging unit and judges; Suspicious tracing object is upgradedModule 16 is in tracing object after more new module 15 is upgraded, then travels through described suspicious tracing object storage mouldEach suspicious tracing object in piece, if there is moving target associated with it, upgrades this suspicious trackingThe life cycle of object, and whether reach the second default existence the life cycle that judges this suspicious tracing objectCt value, is defined as tracing object if reach by this suspicious tracing object, simultaneously by right this suspicious trackingThe property parameters of elephant is updated to the property parameters of the moving target being associated, for there not being the motion being associatedThe suspicious tracing object of target is deleted.
Further, first the first relating module 13 and the second relating module 14 all judge moving target and chase afterTrack object, or whether moving target is crossing with the boundary rectangle of suspicious tracing object, when judged result is for intersecting,Whether the crossing area of further judgement reaches default crossing area threshold, moving target and tracing objectThe area change rate of the boundary rectangle of the area change rate of boundary rectangle or moving target and suspicious tracing object isNo within the scope of default rate of change; If judged result be intersect area reach default crossing area threshold,The area change rate of boundary rectangle, within the scope of default rate of change, judges moving target and tracing objectMatch, or the matching of moving target and suspicious tracing object.
Further, moving object detection unit 1 also comprises moving target extraction module and/or a moving targetProcessing module, described moving target extraction module is crossed the mode that three-frame difference and mixed Gauss model combine and is existedIn the video sequence that camera is taken, extract moving target; Described moving target processing module to extract want chase afterTrack movement destination image carries out morphology processing, so that extract the imperfect areas of disconnection UNICOM causing.
In sum, the present invention utilizes the characteristic of abandon, first abandon is detected and is divided into doubtful abandonStage, preliminary judgement abandon stage and confirmation abandon stage three phases, can avoid false alarm, makes inspectionSurvey precision is improved, and in addition, whole testing process, taking time index and range index as basis for estimation, is calculatedMethod is simple and easy to realize, also can select only the moving target entering completely in scene to be carried out to trace detection simultaneously,And ignore from the interference of the moving object of video edge process, contribute to provide the anti-of whole trace detection computingInterference performance. Above-mentioned abandon detection method is easy to realize, and stable performance, for improving monitoring system to prisonThe understandability of control scene, the security tool that promotes public place plays a very important role. Specifically can be applicable toIn following scene:
(1) be applied to intelligent video monitoring field, for example, when an object is (as chest, parcel, vehicleDeng) overlong time that stops in sensitizing range, or exceeded predefined time span and just produce and report to the police. Allusion quotationType application scenarios comprises airport, railway station, subway station etc., thereby can avoid making because of possible explosive incidentThe massive losses becoming.
(2) be applied to vision guided navigation field, effectively identify and of short durationly occur target and occur for a long time target, therebyThe impact of evaluating objects on environment and travel line in time, realizes the self adaptation in complicated changing environmentNavigation.
The foregoing is only preferred embodiment of the present invention, in order to limit the present invention, not all at thisAny amendment of doing within bright spirit and principle, be equal to and replace and improvement etc., all should be included in the present inventionProtection domain within.
Claims (13)
1. an abandon detection method, is characterized in that, comprises the steps:
Steps A, extracts moving target and follow the trail of at camera in the video sequence of taking;
Described steps A is followed the trail of the wish tracing movement target of extracting by following step:
Steps A 1, by do not reach life cycle default second life cycle threshold value moving target be defined as canDoubt tracing object;
Steps A 2, by reach life cycle default second life cycle threshold value moving target be defined as trackingObject;
Steps A 3, for the wish tracing movement target of extracting in each frame of video sequence, first judgement is fixedIn the tracing object that justice is crossed, whether there is the tracing object matching with it, if the tracing object having definedThe tracing object that middle existence matches with it, is associated this moving target with the tracing object matching;
Steps A 4, if steps A 3 judged results are do not exist in the tracing object having defined and extractWant the tracing object that tracing movement target matches, the suspicious tracing object that further judgement had definedIn whether there is the suspicious tracing object matching with the wish tracing movement target of extracting, if exist, by thisMoving target is associated with the suspicious tracing object matching, if do not exist, this moving target is defined asNew suspicious tracing object;
Steps A 5, each tracing object that first traversal defined, if there is moving target associated with it,Upgrade the life cycle of this tracing object, and the property parameters of the moving target being associated is upgraded and chased after for this reasonThe property parameters of track object; For the tracing object that does not have the moving target being associated, trigger execution stepB；
Steps A 6, after execution step A5, then travels through the each suspicious tracing object defining, if existMoving target associated with it, upgrades the life cycle of this suspicious tracing object, and judges this suspicious chasing afterWhether reach the life cycle of track object default second life cycle threshold value, if reach, this can be followed the trail ofObject definition is tracing object, the property parameters of this suspicious tracing object is updated to the motion being associated simultaneouslyThe property parameters of target, deletes for the suspicious tracing object that does not have the moving target being associated;
Step B, for the moving target of following the trail of in steps A, if there is displacement to be within a predetermined period of time less thanDefault displacement threshold value, and reach its life cycle default first life cycle threshold value, area be greater than defaultThe moving target of area threshold, this moving target is judged as to static doubtful abandon, and records thisThe position of moving target and colouring information;
Step C, for the moving target of following the trail of in steps A, if having moving target current video disappear andThis moving target belongs to the doubtful abandon of step B judgement, by the face of this moving target of step B recordIn look information and current video, the colouring information at same position place mates, if matching degree reaches defaultDegree of joining threshold value, tentatively confirms that this moving target is abandon;
Step D, if the time that the preliminary abandon of confirming of step C stops in video sequence exceedes defaultTime of staying threshold value, reports to the police.
2. abandon detection method as claimed in claim 1, is characterized in that, described step C also wrapsDraw together following step:
Step C1, for the moving target of following the trail of in steps A, if there is moving target to disappear at current videoAnd this moving target does not belong to the doubtful abandon of step B judgement, abandons following the trail of this moving target; Or
Step C2, for the moving target of following the trail of in steps A, if there is moving target to disappear at current videoAnd this moving target belongs to the doubtful abandon of step B judgement, this moving target step B being recordedIn colouring information and current video, the colouring information at same position place mates, default if matching degree does not reachMatching degree threshold value, abandon following the trail of this moving target.
3. abandon detection method as claimed in claim 1, is characterized in that, described step D is concreteComprise the steps:
Step D1, for the preliminary abandon of confirming of step C, by this abandon is remembered in step BIn the colouring information of record and current video, the colouring information at same position place mates, pre-if matching degree reachesIf matching degree threshold value, confirm that this abandon is still present in current video;
Step D2, what statistic procedure D1 confirmed is still present in abandon in current video in video sequenceWhether the time stopping exceedes default time of staying threshold value, reports to the police if exceed.
4. abandon detection method as claimed in claim 3, is characterized in that, described step D is concreteComprise the steps:
Step D3, for the preliminary abandon of confirming of step C, by this abandon is remembered in step BIn the colouring information of record and current video, the colouring information at same position place mates, if matching degree does not reachDefault matching degree threshold value, abandons following the trail of this abandon.
5. abandon detection method as claimed in claim 1, is characterized in that, in described steps A 3, sentencesIn the disconnected tracing object having defined, whether exist the tracking matching with the wish tracing movement target of extracting rightIn the suspicious tracing object resemble, judgement having defined in described steps A 4, whether exist with wanting of extracting and chase afterWhen suspicious tracing object that track moving target matches, all judge by following steps:
Steps A 7, judges moving target and tracing object, or the external square of moving target and suspicious tracing objectWhether shape intersects;
Steps A 8, if steps A 7 judged results are for intersecting, further whether the crossing area of judgement reaches pre-If crossing area threshold, i.e. the area change rate of the boundary rectangle of moving target and tracing object or motion orderWhether the area change rate of mark and the boundary rectangle of suspicious tracing object is within the scope of default rate of change;
Steps A 9, if steps A 8 judged results are to intersect area to reach default crossing area threshold, externalThe area change rate of rectangle, within the scope of default rate of change, judges that moving target and tracing object match,Or moving target and suspicious tracing object match.
6. abandon detection method as claimed in claim 1, is characterized in that, described steps A is passed throughFollowing step is extracted moving target:
Steps A 10, what the mode combining by three-frame difference and mixed Gauss model was taken at camera looksIn frequency sequence, extract moving target;
And/or,
Steps A 11, carries out morphology processing to the wish tracing movement target image extracting, so that cannot extract completelyThe whole areas of disconnection UNICOM causing.
7. an abandon detection system, is characterized in that, comprising:
Moving object detection unit, extracts moving target and follows the trail of for the video sequence of taking at camera;
Described moving object detection unit comprises:
Suspicious tracing object memory module, for store do not reach life cycle default second life cycle thresholdThe moving target of value;
Tracing object memory module, for store reach life cycle default second life cycle threshold value fortuneMoving-target;
The first relating module, for the wish tracing movement target that each frame of video sequence is extracted, first sentencesIn tracing object in disconnected tracing object memory module, whether there is the tracing object matching with it, if existThe tracing object matching with it, is associated this moving target with the tracing object matching;
The second relating module, for being described tracing object storage mould in described the first relating module judged resultThe tracing object that the wish tracing movement target not existing in piece and extract matches, further judgement is described canWhether doubt in tracing object memory module exists the suspicious tracking matching with the wish tracing movement target of extracting rightResemble, if exist, this moving target is associated with the suspicious tracing object matching, if do not exist,This moving target is stored to described suspicious tracing object memory module;
Tracing object is new module more, right for first traveling through each tracking of described tracing object memory moduleResemble, if there is moving target associated with it, the life cycle of upgrading this tracing object, and will be correlated withThe property parameters of the moving target of connection upgrades the property parameters of tracing object for this reason; Be associated for not existingThe tracing object of moving target, triggers doubtful abandon judging unit and judges;
More new module of suspicious tracing object, after in described tracing object, more new module is upgraded, then timeGo through the each suspicious tracing object in described suspicious tracing object memory module, if there is fortune associated with itMoving-target, upgrades the life cycle of this suspicious tracing object, and judges the existence week of this suspicious tracing objectPhase whether reach default second life cycle threshold value, if reach, this suspicious tracing object is defined as to trackingObject is updated to the property parameters of this suspicious tracing object the property parameters of the moving target being associated simultaneously,Delete for the suspicious tracing object that does not have the moving target being associated;
Doubtful abandon judging unit, for the moving target for described moving object detection cell tracks,Judge whether that displacement is within a predetermined period of time less than default displacement threshold value, and reach its life cycle pre-If first life cycle threshold value, area be greater than the moving target of default area threshold, if having by this fortuneMoving-target is judged as static doubtful abandon, and records position and the colouring information of this moving target;
Abandon preliminary judgement unit, for the moving target for described moving object detection cell tracks,Judge whether that moving target belongs to described doubtful abandon judgement list in current video disappearance and this moving targetThe doubtful abandon of unit's judgement, this moving target described doubtful abandon judging unit being recorded if havingIn colouring information and current video, the colouring information at same position place mates, default if matching degree reachesMatching degree threshold value, tentatively confirms that this moving target is abandon;
Abandon determining unit, is looking for the abandon of the preliminary confirmation in described abandon preliminary judgement unitThe time stopping in frequency sequence exceedes default time of staying when threshold value, reports to the police.
8. abandon detection system as claimed in claim 7, is characterized in that, described abandon is tentatively sentencedOrder unit is also for being that the moving target of described moving object detection cell tracks is at current video in judged resultDisappear, and this moving target is not while belonging to the doubtful abandon of described doubtful abandon judging unit judgement controlMake described moving object detection unit and abandon following the trail of this moving target.
9. abandon detection system as claimed in claim 7, is characterized in that, described abandon is tentatively sentencedOrder unit is also for being that the moving target of described moving object detection cell tracks is at current video in judged resultDisappear, and this moving target is while belonging to the doubtful abandon of described doubtful abandon judging unit judgement, by instituteState same position place in the colouring information of this moving target of doubtful abandon judging unit record and current videoColouring information mate, if matching degree does not reach default matching degree threshold value, control described moving targetDetecting unit abandons following the trail of this moving target, and deletes this doubtful abandon simultaneously.
10. abandon detection system as claimed in claim 7, is characterized in that, described abandon is determinedUnit comprises:
The first matching module, for by the preliminary described abandon preliminary judgement unit abandon of confirming by instituteThe colouring information of stating same position place in the colouring information of doubtful abandon judging unit record and current video entersRow coupling, if matching degree reaches default matching degree threshold value, confirms that this abandon is still present in current videoIn;
Alarm module, the abandonment that is still present in current video of confirming for adding up described the first matching moduleWhether the time that thing stops in video sequence exceedes default time of staying threshold value, reports to the police if exceed.
11. abandon detection systems as claimed in claim 10, is characterized in that, when described the first couplingResult after module coupling is matching degree while not reaching default matching degree threshold value, and described the first matching module alsoBe used for controlling described moving object detection unit and abandon following the trail of this moving target, delete this doubtful abandon simultaneously.
12. abandon detection systems as claimed in claim 7, is characterized in that, described first closes gang mouldFirst piece and the second relating module all judge moving target and tracing object, or moving target and suspicious tracking rightWhether the boundary rectangle of elephant intersects, and when judged result is for intersecting, further judges whether intersect area reachesArea change rate or the motion order of the boundary rectangle of default crossing area threshold, moving target and tracing objectWhether the area change rate of mark and the boundary rectangle of suspicious tracing object is within the scope of default rate of change; If sentenceDisconnected result is that crossing area reaches default crossing area threshold, the area change rate of boundary rectangle is being presetWithin the scope of rate of change, judge matching of moving target and tracing object, or moving target and suspicious trackingMatching of object.
13. abandon detection systems as claimed in claim 7, is characterized in that, described moving target inspectionMeasurement unit comprises:
Moving target extraction module, is taking the photograph for the mode combining by three-frame difference and mixed Gauss modelIn the video sequence that picture head is taken, extract moving target;
And/or,
Moving target processing module, for the wish tracing movement target image extracting is carried out to morphology processing,So that extract the imperfect areas of disconnection UNICOM causing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210180275.9A CN103455997B (en) | 2012-06-04 | 2012-06-04 | A kind of abandon detection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210180275.9A CN103455997B (en) | 2012-06-04 | 2012-06-04 | A kind of abandon detection method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103455997A CN103455997A (en) | 2013-12-18 |
| CN103455997B true CN103455997B (en) | 2016-05-04 |
Family
ID=49738327
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210180275.9A Expired - Fee Related CN103455997B (en) | 2012-06-04 | 2012-06-04 | A kind of abandon detection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103455997B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105740814B (en) * | 2016-01-29 | 2018-10-26 | 重庆扬讯软件技术股份有限公司 | A method of determining solid waste dangerous waste storage configuration using video analysis |
| CN108629254B (en) * | 2017-03-24 | 2020-11-27 | 杭州海康威视数字技术股份有限公司 | Moving target detection method and device |
| CN107145861A (en) * | 2017-05-05 | 2017-09-08 | 中国科学院上海高等研究院 | A kind of abandon automatic testing method |
| CN110443834A (en) * | 2018-05-04 | 2019-11-12 | 大猩猩科技股份有限公司 | A kind of distributed object tracking system |
| CN111435418B (en) * | 2018-12-26 | 2024-01-02 | 深圳市优必选科技有限公司 | Method and device for identifying personalized object of robot, storage medium and robot |
| CN114399537B (en) * | 2022-03-23 | 2022-07-01 | 东莞先知大数据有限公司 | Vehicle tracking method and system for target personnel |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004015603A (en) * | 2002-06-10 | 2004-01-15 | Seiwa Electric Mfg Co Ltd | Object detection system |
| CN101231696A (en) * | 2008-01-30 | 2008-07-30 | 安防科技（中国）有限公司 | Method and system for detection of hangover |
| CN101635026A (en) * | 2008-07-23 | 2010-01-27 | 中国科学院自动化研究所 | Method for detecting derelict without tracking process |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8284249B2 (en) * | 2008-03-25 | 2012-10-09 | International Business Machines Corporation | Real time processing of video frames for triggering an alert |
-
2012
- 2012-06-04 CN CN201210180275.9A patent/CN103455997B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004015603A (en) * | 2002-06-10 | 2004-01-15 | Seiwa Electric Mfg Co Ltd | Object detection system |
| CN101231696A (en) * | 2008-01-30 | 2008-07-30 | 安防科技（中国）有限公司 | Method and system for detection of hangover |
| CN101635026A (en) * | 2008-07-23 | 2010-01-27 | 中国科学院自动化研究所 | Method for detecting derelict without tracking process |
Non-Patent Citations (2)
| Title |
|---|
| 基于全方位视觉的遗留物及其放置者检测的研究;富吉勇;《万方学位论文数据库》;20111031;第16、23-24、39-40、48页，图3-1、3-3 * |
| 混合高斯模型背景法的一种改进算法;刘静 等;《计算机工程与应用》;20100501;第168页摘要，第170页第3.3.2节 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103455997A (en) | 2013-12-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103455997B (en) | A kind of abandon detection method and system | |
| Rezazadeh Azar et al. | Server-customer interaction tracker: computer vision–based system to estimate dirt-loading cycles | |
| Yoneyama et al. | Moving cast shadow elimination for robust vehicle extraction based on 2D joint vehicle/shadow models | |
| CN101577006B (en) | Loitering detecting method and loitering detecting system in video monitoring | |
| CN101295405A (en) | Portrait and vehicle recognition alarming and tracing method | |
| CN101393609B (en) | Target detection tracking method and device | |
| CN105744232A (en) | Method for preventing power transmission line from being externally broken through video based on behaviour analysis technology | |
| Jodoin et al. | Tracking all road users at multimodal urban traffic intersections | |
| CN102496281B (en) | Vehicle red-light violation detection method based on combination of tracking and virtual loop | |
| CN104200490A (en) | Rapid retrograde detecting and tracking monitoring method under complex environment | |
| CN104063885A (en) | Improved movement target detecting and tracking method | |
| Stauffer | Learning to track objects through unobserved regions | |
| CN105513349A (en) | Double-perspective learning-based mountainous area highway vehicle event detection method | |
| CN103280052B (en) | Be applied to the intrusion detection method of long distance track circuit intelligent video monitoring | |
| CN104408406A (en) | Staff off-post detection method based on frame difference method and background subtraction method | |
| CN106228572B (en) | A kind of the long inactivity object detection and tracking of carrier state mark | |
| CN105575125A (en) | Vehicle flow video detection and analysis system | |
| CN103729861A (en) | Multiple object tracking method | |
| Nienhüser et al. | Visual state estimation of traffic lights using hidden Markov models | |
| Cuevas et al. | Detection of stationary foreground objects using multiple nonparametric background-foreground models on a finite state machine | |
| CN103985257A (en) | Intelligent traffic video analysis method | |
| CN102831378A (en) | Method and system for detecting and tracking human object | |
| CN109872541A (en) | A kind of information of vehicles analysis method and device | |
| CN104168444A (en) | Target tracking method of tracking ball machine and tracking ball machine | |
| CN112560641A (en) | Video-based one-way passenger flow information detection method in two-way passenger flow channel |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20160504 Termination date: 20210604 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |