CN107644206A - A kind of road abnormal behaviour action detection device - Google Patents
A kind of road abnormal behaviour action detection device Download PDFInfo
- Publication number
- CN107644206A CN107644206A CN201710855428.8A CN201710855428A CN107644206A CN 107644206 A CN107644206 A CN 107644206A CN 201710855428 A CN201710855428 A CN 201710855428A CN 107644206 A CN107644206 A CN 107644206A
- Authority
- CN
- China
- Prior art keywords
- track
- target
- representing
- video
- tracking 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
- 206010000117 Abnormal behaviour Diseases 0.000 title claims abstract description 43
- 238000001514 detection method Methods 0.000 title claims abstract description 17
- 230000009471 action Effects 0.000 title abstract description 5
- 230000033001 locomotion Effects 0.000 claims abstract description 31
- 238000012545 processing Methods 0.000 claims abstract description 13
- 238000004891 communication Methods 0.000 claims abstract description 5
- 239000013598 vector Substances 0.000 claims description 26
- 239000011159 matrix material Substances 0.000 claims description 19
- 230000006870 function Effects 0.000 claims description 18
- 238000000034 method Methods 0.000 claims description 13
- 239000006185 dispersion Substances 0.000 claims description 10
- 230000005484 gravity Effects 0.000 claims description 10
- 238000012544 monitoring process Methods 0.000 claims description 6
- 238000004458 analytical method Methods 0.000 claims description 4
- 230000001186 cumulative effect Effects 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 4
- 230000006399 behavior Effects 0.000 claims description 3
- 230000008859 change Effects 0.000 claims description 2
- 238000000354 decomposition reaction Methods 0.000 claims description 2
- 230000008569 process Effects 0.000 claims description 2
- 238000012163 sequencing technique Methods 0.000 claims description 2
- 239000000284 extract Substances 0.000 abstract 1
- 230000002159 abnormal effect Effects 0.000 description 4
- 206010039203 Road traffic accident Diseases 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000008439 repair process Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000033228 biological regulation Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000006698 induction Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
A kind of road abnormal behaviour action detection device, including：Video acquisition module, it is connected with video identification module, for obtaining the original video of road vehicles motion；Video identification module, for realizing video processnig algorithms, the abnormal behaviour of vehicle is detected in real time and extracts the license board information of abnormal behaviour vehicle；Video processing module, it is connected with the identification module, for the license board information of the abnormal behaviour vehicle of acquisition to be embedded in video information, the video after being handled；Communication module, for being sent the video after original video and processing to administrative center by wireless network.The present invention can in automatic identification current video picture vehicle abnormal behaviour, real-time and recognition accuracy is high.
Description
Technical Field
The invention relates to the field of road real-time detection, in particular to a road abnormal behavior action detection device.
Background
Along with the development of economy and the improvement of people's living standard, people's daily trip is more and more frequent, and urban traffic is more and more crowded, and traffic accident also takes place occasionally, and most of traffic accidents are because vehicle or pedestrian violate the rule and cause, and the traffic violation does not nevertheless influence the traffic road surface's passage, but also brings important potential safety hazard to people's the lives and properties, consequently how to stop the traffic violation phenomenon and become the hot problem that the traffic department waited to solve urgently.
At present, a general traffic monitoring system determines violation behaviors of vehicles by using an electronic camera shooting supervision technology at important road sections such as a crossroad and the like, and the main limitation of the mode is that: the license plate of the vehicle must be identified through a complex algorithm to determine the offending vehicle, and the accuracy and reliability thereof are limited by a plurality of factors. If at night, when the weather is bad and the speed of a motor vehicle is too high, the definition of a shot image is low, and when shielding is generated among vehicles, even the license plate of a vehicle violating the regulations cannot be shot. Therefore, under the current conditions, traffic control departments mainly rely on identifying the violation vehicles through manual methods (watching traffic videos), which requires a large amount of manpower and means that the violation judgment efficiency and the real-time performance are not high. Another technique is to bury a sensor underground, which is not restricted by road visibility, but an underground induction line is easily damaged by heavy vehicles, road surface repair, etc., and repair or installation of the sensor interrupts traffic and affects road surface life, and thus affects the smoothness of traffic to a certain extent. In a word, the existing technical means cannot well solve the problem of traffic violation in large quantity.
Disclosure of Invention
In view of the above problems, the present invention is directed to a road abnormal behavior detection device.
The purpose of the invention is realized by adopting the following technical scheme:
a road abnormal behavior action detection device includes:
the video acquisition module is connected with the video identification module and used for acquiring an original video of the movement of the vehicle on the road surface;
the video recognition module is used for realizing a video processing algorithm, detecting abnormal behaviors of the vehicle in real time and extracting license plate information of the vehicle with the abnormal behaviors;
the video processing module is connected with the identification module and used for embedding the acquired license plate information of the abnormal behavior vehicle into the video information to obtain a processed video;
and the communication module is used for sending the original video and the processed video to the management center through a wireless network.
The video acquisition module is a network monitoring camera and is arranged at a position with a certain height and a wide angle.
Wherein the video identification module comprises:
the moving vehicle detection unit is used for extracting a background image and a foreground image in an original video, performing motion recognition and obtaining a moving vehicle in the foreground image as a tracking target;
the moving vehicle tracking unit is used for acquiring a motion track of a tracking target;
the vehicle track learning unit is used for establishing a standard track model according to the collected vehicle track sample data in a learning stage;
the abnormal behavior detection unit is used for judging whether the target track has abnormal behaviors or not by taking the motion track of the tracking target as the target track;
license plate information extraction unit: the method is used for extracting the license plate information of the vehicle with the abnormal behavior in real time.
The invention has the beneficial effects that: the invention can acquire the motion information of all vehicles on the road, quickly judge whether the vehicles have abnormal behaviors or not, and send the videos of the vehicles with the abnormal behaviors and the license plate information to the management center in real time, thereby effectively and quickly reflecting the road abnormal conditions to the management center, providing guarantee for a manager to immediately take measures and improving the intelligent level of road monitoring.
Drawings
The invention is further illustrated by means of the attached drawings, but the embodiments in the drawings do not constitute any limitation to the invention, and further drawings may be obtained by those skilled in the art without inventive effort, based on the following drawings.
FIG. 1 is a block diagram of the frame of the present invention;
fig. 2 is a block diagram of a frame of a video recognition module according to the present invention.
Reference numerals:
the system comprises a video acquisition module 1, a video recognition module 2, a video processing module 3, a communication module 4, a moving vehicle detection module 20, a moving vehicle tracking module 21, a vehicle track learning module 22, an abnormal behavior detection module 23 and a license plate information extraction unit 24
Detailed Description
The invention is further described in connection with the following application scenarios.
A road abnormal behavior detection apparatus, see fig. 1, comprising:
the video acquisition module 1 is connected with the video identification module 2 and is used for acquiring an original video of the movement of the vehicle on the road surface;
the video identification module 2 is used for realizing a video processing algorithm, detecting abnormal behaviors of the vehicle in real time and extracting license plate information of the abnormal behavior vehicle;
the video processing module 3 is connected with the recognition module 2 and used for embedding the acquired license plate information of the abnormal behavior vehicle into the video information to obtain a processed video;
and the communication module 4 is used for sending the original video and the processed video to a management center through a wireless network.
Preferably, the video capture module 1 is a network monitoring camera, and is disposed at a position with a certain height and a wide angle.
Preferably, referring to fig. 2, the video identification module 2 comprises:
moving vehicle detection unit 20: the method comprises the steps of extracting a background image and a foreground image in an original video, performing motion recognition, and acquiring a moving vehicle in the foreground image as a tracking target;
a moving vehicle tracking unit 21 for acquiring a movement locus of a tracking target;
the vehicle track learning unit 22 is used for establishing a standard track model according to the collected vehicle track sample data in a learning stage;
an abnormal behavior detection unit 23, which takes the motion trajectory of the tracking target as a target trajectory, and determines whether an abnormal behavior exists in the target trajectory;
and the license plate information extraction unit 24 is used for extracting the license plate information of the vehicle with the abnormal behavior in real time.
According to the embodiment of the invention, the motion information of all vehicles on the road can be obtained, whether the vehicles have abnormal behaviors or not can be quickly judged, the video and the license plate information of the vehicles with the abnormal behaviors are sent to the management center in real time, the abnormal conditions of the road can be effectively and quickly reflected to the management center, the guarantee is provided for a manager to take measures immediately, and the intelligent level of road monitoring is improved.
Preferably, the moving vehicle tracking unit 21 is configured to acquire a motion trajectory of a tracking target, and includes:
(1) establishing a tracking target state description model for describing the state characteristics of a tracking target R:
Xk＝(Ck,Vk,Mk,Gk)
in the formula, XkRepresenting the state feature vector of the tracked object at time k, CkIndicating the position of the center of gravity, V, of the tracked object at time kkRepresenting the velocity of movement, V, of the tracked object at time kk＝Ck-Ck-1，MkRepresenting the eccentricity vector dispersion of the tracked target at time k,wherein k ismRepresents the total dimension of the eccentric moment vector, D' (i) represents the normalized vector of the eccentric moment of the ith dimension,representing mean value of eccentricity vectors, GkRepresenting the gray scale feature of the tracked object at time k,wherein n represents the total number of pixel points in the tracking target R, and I (I, j) represents the gray value of any pixel point in the tracking target R;
(2) estimating the state characteristics of the tracking target, wherein the adopted state characteristic estimation function is as follows:
Ck＝Ck-1+Ck-1×Δt+μ1
Gk＝Gk-1+μ3
in the formula, CkAnd Ck-1Respectively representing the position of the center of gravity of the tracked target at the time k and the time k-1, delta t representing the time interval between adjacent observation times, mu1Indicating a position estimation error, MkAnd Mk-1Respectively represents the decentration of the eccentric moment vector of the tracking target at the moment k and the moment k-1,represents the variation of the vector dispersion degree of the tracking target eccentric moment predicted by the moment k-1 to the next moment, gamma denotes an update factor which is a function of,represents the variation of the vector dispersion degree mu of the tracked target eccentricity predicted from the k-2 moment to the next moment2Representing the estimated error of the dispersion of the moment of eccentricity vectors, GkAnd Gk-1Representing the gray features, mu, of the tracked object at times k and k-1, respectively3Representing a gray level feature estimation error;
(3) matching the state characteristic component estimated value of the tracked target at the previous moment with the state characteristic components of all unmatched foreground targets at the current moment, namely matching the state characteristic components of the gravity center position, the dispersion degree of the eccentric moment vector, the gray degree and the like of all the unmatched foreground targets with the tracked targets one by one, if the distance of a certain state characteristic component is smaller than a set threshold value, determining that the state characteristic component is successfully matched, and if two or more than two characteristics of one unmatched foreground target are successfully matched, matching the unmatched foreground target as the tracked target; otherwise, when all unmatched foreground targets have only one or no feature component, the matching is successful, which may be caused by the occlusion situation, and the occlusion processing analysis needs to be performed on the tracked target;
(4) the occlusion processing analysis comprises the following steps: when the shielding situation occurs, predicting the state characteristics, specifically:
predicting the state characteristics of the tracking target by using a custom state characteristic prediction model:
wherein,
in the formula,representing the prediction of the state characteristics of the tracked target at time k +1, u0(1) Initial state features representing the tracking targets, α and β represent state feature prediction parameters, a and b represent parameters to be estimated, and (a, b) is satisfiedT＝(BTB)-1BTU, where B represents the cumulative state signature sequence,u represents a sequence of state characteristics that are,u1(n) represents the cumulative state characteristics of the tracked target at time n, u0(n) represents the state characteristics of the tracking target at the time n,
if the tracking target at the current kth moment is not matched with all foreground targets, considering that the tracking target is possibly shielded, temporarily retaining the unmatched tracking target, marking the unmatched tracking target, establishing a temporary shielding linked list, adding the historical state features of the tracking target into the temporary shielding linked list, simultaneously updating the target state by adopting the custom state feature prediction model, predicting the motion state of the shielding process, if the tracking target is successfully matched with the foreground targets again within T moments, considering that the tracking target is temporarily shielded, returning the tracking target to a normal tracking state, and if the tracking target is still not successfully matched within T moments, considering that the tracking target disappears, wherein T is a set shielding time threshold value;
(5) and recording the change of the gravity center position of the tracking target as a motion track of the tracking target.
According to the preferred embodiment, the vehicle is tracked by adopting the method, the motion track of the vehicle can be accurately estimated and matched according to the characteristic parameters of the vehicle, the motion track of the vehicle is effectively obtained, and the adaptability is high; particularly, under the condition that the vehicle is possibly shielded, the motion state of the shielded vehicle can be accurately simulated by adopting the customized vehicle state characteristic prediction model, the accuracy of vehicle track acquisition is effectively improved, and a foundation is laid for the subsequent judgment of abnormal behavior of the vehicle.
Preferably, the vehicle track learning unit 22 is configured to, in the learning phase, establish a standard track model according to the collected vehicle track sample data, and includes:
(1) the method comprises the steps that a motion track of a normal behavior of a vehicle is collected and screened out through a video collection module to serve as an effective sample track;
(2) for a sample T comprising k valid sample tracks T ═ T1,T2,…,TkAnd establishing a user-defined similar matrix S:
wherein,
s(Ti,Tj)＝exp(-H(Ti,Tj)/2σ2)
H(Ti,Tj)＝min(h(Ti,Tj),h(Tj,Ti))
in the formula, Si,jRepresents the value of the ith row and the jth column in the similarity matrix, TiAnd TjRespectively representing the ith and jth effective sample tracks in the sample, i, j E [1, k ∈ k]K denotes the total number of valid sample tracks, s (T)i,Tj) Representing valid sample tracks TiAnd TjThe self-defined similarity function is used for calculating the effective sample track TiAnd TjSimilarity of (D), H (T)i,Tj) Representing the effective track TiAnd TjA symmetric distance between the target track a and the track b, a represents a scale parameter, h (a, b) represents a directed distance function for calculating the directed distance between the target track a and the track b, NrRepresenting the number of trace points of the target trajectory in the directed distance function,andrespectively represent the track TiThe abscissa and ordinate of the mth track point,andrespectively represented in the valid sample track TjMiddle and valid sample trajectory TjThe abscissa and ordinate of the trace point closest to the mth trace point in the middle,representing valid sample tracks TjMiddle and valid sample trajectory TjThe sequencing of the trace point with the mth trace point nearest to the mth trace point, andrespectively representing valid sample tracks TjThe abscissa and the ordinate of the nth track point, wherein n is less than or equal to the effective sample track TjThe total number of the middle trace points,the function represents the value of n when the coincidence function A (n) is taken as the minimum value;
(3) constructing a Laplace matrix L according to the similarity matrix S and the degree matrix D, wherein L ═ D-1/2SD-1/2，Wherein Si,jRepresenting the value of the ith row and the jth column in the similar matrix, wherein k represents the total number of effective sample tracks;
(4) performing characteristic value decomposition on L by using a Krylov subspace iteration method;
(5) all the characteristic values lambda are sorted from small to large1≤λ2≤…≤λkIf the difference between the b-th eigenvalue and the b + 1-th eigenvalue is large, that isB is taken as the number of the clustering categories;
(6) b minimum eigenvalues, namely corresponding eigenvectors u are calculated1,u2,…,ubConstruct k × b matrix U ═ U1,u2,…,ub](ii) a Normalizing each line of the U to obtain a matrix U';
(7) consider the row vector of U' as a point in b-dimensional space, each track sample TiAnd clustering the ith row vector corresponding to the U' in a space by using a k-means clustering algorithm, and dividing the track sample into b types as a set b type track mode.
In the preferred embodiment, the sample tracks are classified and learned by the method, the similarity among the sample tracks can be effectively and accurately described by self-defining the similarity function, then the similar matrix is established according to the similarity among the sample tracks, and each sample track is accurately and reasonably classified by the clustering algorithm, so that the accurate judgment of the target abnormal behavior is guaranteed.
Preferably, the abnormal behavior detecting unit 23, taking the motion trajectory of the tracking target as a target trajectory, and determining whether there is an abnormal behavior in the target trajectory, includes:
(1) setting the track mode of the sample track divided into b types as omega1,ω2,…,ωbJudging the mode attribution of each track point in the target track A by adopting a custom track mode attribution function:
in the formula, pr(at) The track point in the target track at the moment t belongs to the track mode omegarR 1,2, …, b, b represents the total number of track pattern classes, atRepresenting the track points in the target track at time t,represents the track pattern ωrTwo-dimensional mean vector of middle samples, SrRepresenting a two-dimensional covariance matrix, P (ω)r) Represents the track pattern ωrThe occupied specific gravity;
(2) sequentially obtaining the attribution value of each track point in the target track A to each track mode, if the maximum attribution value of each track point in the target track A belongs to the same track mode and the attribution value is larger than a set threshold value, namelyAnd p isδ(at)>θjThen, the target track is judged to belong to the track mode omegaδ(ii) a Conversely, if for the track mode ω1,ω2,…,ωbIn any mode, if the target track A does not meet the condition, judging that the target track A has abnormal behavior;
wherein p isr(at) Showing that track points in the target track at the t-th moment belong to a track mode omegarR 1,2, …, b, b represents the total number of track pattern classes, atShowing track points in the target track at the t-th moment, and delta showing that the track points in the target track at the t-th moment belong to a track mode omegaδMaximum ascribed value of thetajIndicating a set home threshold.
In the preferred embodiment, the mode attribution judgment is performed on each track point in the target track by adopting the custom track mode attribution function, and whether the target track is abnormal or not can be judged quickly and accurately by judging the mode attribution of each track point in the target track, so that the method is suitable for judging the road abnormal condition in real time and provides a basis for a management center to make a response measure quickly.
Finally, it should be noted that the above embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the protection scope of the present invention, although the present invention is described in detail with reference to the preferred embodiments, it should be analyzed by those skilled in the art that modifications or equivalent substitutions can be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.
Claims (6)
1. A road abnormal behavior motion detection device is characterized by comprising:
the video acquisition module is connected with the video identification module and used for acquiring an original video of the movement of the vehicle on the road surface;
the video recognition module is used for realizing a video processing algorithm, detecting abnormal behaviors of the vehicle in real time and extracting license plate information of the vehicle with the abnormal behaviors;
the video processing module is connected with the identification module and used for embedding the acquired license plate information of the abnormal behavior vehicle into the video information to obtain a processed video;
and the communication module is used for sending the original video and the processed video to the management center through a wireless network.
2. The device of claim 1, wherein the video capture module is a network monitoring camera disposed at a position with a certain height and a wide angle.
3. The device according to claim 1, wherein the video recognition module comprises:
the moving vehicle detection unit is used for extracting a background image and a foreground image in an original video, performing motion recognition and acquiring a moving vehicle in the foreground image as a tracking target;
the moving vehicle tracking unit is used for acquiring a motion track of a tracking target;
the vehicle track learning unit is used for establishing a standard track model according to the collected vehicle track sample data in a learning stage;
the abnormal behavior detection unit is used for judging whether the target track has abnormal behaviors or not by taking the motion track of the tracking target as the target track;
and the license plate information extraction unit is used for extracting the license plate information of the vehicle with the abnormal behavior in real time.
4. The device according to claim 3, wherein the moving vehicle tracking unit is configured to obtain a motion trajectory of a tracking target, and includes:
(1) establishing a tracking target state description model for describing the state characteristics of a tracking target R:
Xk＝(Ck,Vk,Mk,Gk)
in the formula, XkRepresenting the state feature vector of the tracked object at time k, CkIndicating the tracked target at time kPosition of center of gravity of, VkRepresenting the velocity of movement, V, of the tracked object at time kk＝Ck-Ck-1，MkRepresenting the eccentricity vector dispersion of the tracked target at time k,wherein k ismRepresents the total dimension of the eccentric moment vector, D' (i) represents the normalized vector of the eccentric moment of the ith dimension,representing mean value of eccentricity vectors, GkRepresenting the gray scale feature of the tracked object at time k,wherein n represents the total number of pixel points in the tracking target R, and I (I, j) represents the gray value of any pixel point in the tracking target R;
(2) estimating the state characteristics of the tracking target, wherein the adopted state characteristic estimation function is as follows:
Ck＝Ck-1+Ck-1×Δt+μ1
Gk＝Gk-1+μ3
in the formula, CkAnd Ck-1Respectively representing the position of the center of gravity of the tracked target at the time k and the time k-1, delta t representing the time interval between adjacent observation times, mu1Indicating a position estimation error, MkAnd Mk-1Respectively represents the decentration of the eccentric moment vector of the tracking target at the moment k and the moment k-1,represents the variation of the vector dispersion degree of the tracking target eccentric moment predicted by the moment k-1 to the next moment, gamma denotes an update factor which is a function of,represents the variation of the vector dispersion degree mu of the tracked target eccentricity predicted from the k-2 moment to the next moment2Representing the estimated error of the dispersion of the moment of eccentricity vectors, GkAnd Gk-1Representing the gray features, mu, of the tracked object at times k and k-1, respectively3Representing a gray level feature estimation error;
(3) matching the state characteristic component estimated value of the tracked target at the previous moment with the state characteristic components of all unmatched foreground targets at the current moment, namely matching the state characteristic components of the gravity center positions, the dispersion degrees of the eccentric moment vectors, the gray levels and the like of all the unmatched foreground targets with the tracked targets one by one, if the distance of a certain state characteristic component is smaller than a set threshold value, considering that the state characteristic component is successfully matched, and if two or more than two characteristics of one unmatched foreground target are successfully matched, matching the unmatched foreground target as the tracked target; otherwise, when all unmatched foreground targets have only one or no feature component, the matching is successful, which may be caused by the occlusion situation, and the occlusion processing analysis needs to be performed on the tracked target;
(4) the occlusion processing analysis comprises the following steps: when the shielding situation occurs, predicting the state characteristics, specifically:
predicting the state characteristics of the tracking target by using a custom state characteristic prediction model:
wherein,
in the formula,representing the prediction of the state characteristics of the tracked target at time k +1, u0(1) Initial state features representing the tracking targets, α and β representing state feature prediction parameters, a and b representing parameters to be estimated, and (a, b) being satisfiedT＝(BTB)-1BTU, where B represents the cumulative state signature sequence,u represents a sequence of state characteristics that are,u1(n) represents the cumulative state characteristics of the tracked target at time n, u0(n) represents the state characteristics of the tracking target at the time n,
if the tracking target at the current kth moment is not matched with all foreground targets, considering that the tracking target is possibly shielded, temporarily retaining the unmatched tracking target, marking the unmatched tracking target, establishing a temporary shielding linked list, adding the historical state features of the tracking target into the temporary shielding linked list, simultaneously updating the target state by adopting the custom state feature prediction model, predicting the motion state of the shielding process, if the tracking target is successfully matched with the foreground targets again within T moments, considering that the tracking target is temporarily shielded, returning the tracking target to a normal tracking state, and if the tracking target is still not successfully matched within T moments, considering that the tracking target disappears, wherein T is a set shielding time threshold value;
(5) and recording the change of the gravity center position of the tracking target as a motion track of the tracking target.
5. The device according to claim 4, wherein the vehicle track learning unit is configured to establish a standard track model according to the collected vehicle track sample data in a learning phase, and includes:
(1) the method comprises the steps that a motion track of a normal behavior of a vehicle is collected and screened out through a video collection module to serve as an effective sample track;
(2) for a sample T comprising k valid sample tracks T ═ T1,T2,…,TkAnd establishing a user-defined similar matrix S:
wherein,
s(Ti,Tj)＝exp(-H(Ti,Tj)/2σ2)
H(Ti,Tj)＝min(h(Ti,Tj),h(Tj,Ti))
in the formula, Si,jRepresents the value of the ith row and the jth column in the similarity matrix, TiAnd TjRespectively representing the ith and jth effective sample tracks in the sample, i, j E [1, k ∈ k]K denotes the total number of valid sample tracks, s (T)i,Tj) Representing valid sample tracks TiAnd TjThe self-defined similarity function is used for calculating the effective sample track TiAnd TjSimilarity of (D), H (T)i,Tj) Representing the effective track TiAnd TjA symmetric distance between the target track a and the track b, a represents a scale parameter, h (a, b) represents a directed distance function for calculating the directed distance between the target track a and the track b, NrRepresenting the number of trace points of the target trajectory in the directed distance function,andrespectively represent the track TiThe abscissa and ordinate of the mth track point,andrespectively represented in the valid sample track TjMiddle and valid sample trajectory TjThe abscissa and ordinate of the trace point closest to the mth trace point in the middle,representing valid sample tracks TjMiddle and valid sample trajectory TjThe sequencing of the trace point with the mth trace point nearest to the mth trace point, andrespectively representing valid sample tracks TjThe abscissa and the ordinate of the nth track point, wherein n is less than or equal to the effective sample track TjThe total number of the middle trace points,the function represents the value of n when the coincidence function A (n) is taken as the minimum value;
(3) constructing a Laplace matrix L according to the similarity matrix S and the degree matrix D, wherein L ═ D-1/2SD-1/2，Wherein Si,jRepresenting the value of the ith row and the jth column in the similar matrix, wherein k represents the total number of effective sample tracks;
(4) performing characteristic value decomposition on L by using a Krylov subspace iteration method;
(5) all the characteristic values lambda are sorted from small to large1≤λ2≤…≤λkIf the difference between the b-th eigenvalue and the b + 1-th eigenvalue is large, that isB is taken as the number of the clustering categories;
(6) b minimum eigenvalues, namely corresponding eigenvectors u are calculated1,u2,…,ubConstruct k × b matrix U ═ U1,u2,…,ub](ii) a Normalizing each line of the U to obtain a matrix U';
(7) consider the row vector of U' as a point in b-dimensional space, each track sample TiAnd clustering the ith row vector corresponding to the U' in the space by using a clustering algorithm, and dividing the track sample into b types as a set b type track mode.
6. The device according to claim 5, wherein the abnormal behavior detection unit determines whether there is an abnormal behavior in a target track by using a motion track of the tracking target as the target track, and includes:
(1) setting the track mode of the sample track divided into b types as omega1,ω2,…,ωbJudging the mode attribution of each track point in the target track A by adopting a custom track mode attribution function:
in the formula, pr(at) When represents tThe track points in the carved target track belong to a track mode omegarR 1,2, …, b, b represents the total number of track pattern classifications, atRepresenting the track points in the target track at time t,represents the track pattern ωrTwo-dimensional mean vector of middle samples, SrRepresenting a two-dimensional covariance matrix, P (ω)r) Represents the track pattern ωrThe occupied specific gravity;
(2) sequentially obtaining the attribution value of each track point in the target track A to each track mode, if the maximum attribution value of each track point in the target track A belongs to the same track mode and the attribution value is larger than a set threshold value, namelyAnd p isδ(at)>θjThen, the target track is judged to belong to the track mode omegaδ(ii) a Conversely, if for the track mode ω1,ω2,…,ωbIn any mode, if the target track A does not meet the condition, judging that the target track A has abnormal behavior;
wherein p isr(at) Showing that track points in the target track at the t-th moment belong to a track mode omegarR 1,2, …, b, b represents the total number of track pattern classifications, atShowing track points in the target track at the t-th moment, and delta showing that the track points in the target track at the t-th moment belong to a track mode omegaδMaximum ascribed value of thetajIndicating a set home threshold.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710855428.8A CN107644206A (en) | 2017-09-20 | 2017-09-20 | A kind of road abnormal behaviour action detection device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710855428.8A CN107644206A (en) | 2017-09-20 | 2017-09-20 | A kind of road abnormal behaviour action detection device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107644206A true CN107644206A (en) | 2018-01-30 |
Family
ID=61113744
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710855428.8A Pending CN107644206A (en) | 2017-09-20 | 2017-09-20 | A kind of road abnormal behaviour action detection device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107644206A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108806269A (en) * | 2018-06-22 | 2018-11-13 | 安徽科力信息产业有限责任公司 | A kind of method and device of record motor vehicle continuous transformation track illegal activities |
| CN110536114A (en) * | 2019-08-05 | 2019-12-03 | 特斯联（北京）科技有限公司 | A kind of the parking lot CCTV monitoring system and method for Intelligent target tracking |
| CN110728842A (en) * | 2019-10-23 | 2020-01-24 | 江苏智通交通科技有限公司 | Abnormal driving early warning method based on reasonable driving range of vehicles at intersection |
| WO2020057346A1 (en) * | 2018-09-21 | 2020-03-26 | 深圳市九洲电器有限公司 | Video monitoring method and apparatus, monitoring server and video monitoring system |
| CN111105437A (en) * | 2018-10-29 | 2020-05-05 | 西安宇视信息科技有限公司 | Vehicle track abnormity judgment method and device |
| CN111597905A (en) * | 2020-04-20 | 2020-08-28 | 淮阴工学院 | Highway tunnel parking detection method based on video technology |
| CN112164223A (en) * | 2020-02-27 | 2021-01-01 | 周国霞 | Intelligent traffic information processing method and device based on cloud platform |
| CN112673406A (en) * | 2020-05-29 | 2021-04-16 | 华为技术有限公司 | Method and terminal device for identifying abnormal vehicle parameters in vehicle queue |
| CN112700474A (en) * | 2020-12-31 | 2021-04-23 | 广东美的白色家电技术创新中心有限公司 | Collision detection method, device and computer-readable storage medium |
| CN113658423A (en) * | 2020-10-23 | 2021-11-16 | 秦胜君 | Vehicle track abnormity detection method based on circulating gate control unit |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102298844A (en) * | 2011-08-15 | 2011-12-28 | 无锡中星微电子有限公司 | Automatic rule breaking vehicle detection system and method |
| CN105070053A (en) * | 2015-07-21 | 2015-11-18 | 武汉理工大学 | Intelligent traffic monitoring camera for identifying vehicle illegal movement modes |
-
2017
- 2017-09-20 CN CN201710855428.8A patent/CN107644206A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102298844A (en) * | 2011-08-15 | 2011-12-28 | 无锡中星微电子有限公司 | Automatic rule breaking vehicle detection system and method |
| CN105070053A (en) * | 2015-07-21 | 2015-11-18 | 武汉理工大学 | Intelligent traffic monitoring camera for identifying vehicle illegal movement modes |
Non-Patent Citations (1)
| Title |
|---|
| 胡宏宇: "基于视频处理的交通事件识别方法研究", 《中国博士学位论文全文数据库 信息科技辑》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108806269A (en) * | 2018-06-22 | 2018-11-13 | 安徽科力信息产业有限责任公司 | A kind of method and device of record motor vehicle continuous transformation track illegal activities |
| WO2020057346A1 (en) * | 2018-09-21 | 2020-03-26 | 深圳市九洲电器有限公司 | Video monitoring method and apparatus, monitoring server and video monitoring system |
| CN111105437B (en) * | 2018-10-29 | 2024-03-29 | 西安宇视信息科技有限公司 | Vehicle track abnormality judging method and device |
| CN111105437A (en) * | 2018-10-29 | 2020-05-05 | 西安宇视信息科技有限公司 | Vehicle track abnormity judgment method and device |
| CN110536114A (en) * | 2019-08-05 | 2019-12-03 | 特斯联（北京）科技有限公司 | A kind of the parking lot CCTV monitoring system and method for Intelligent target tracking |
| CN110728842B (en) * | 2019-10-23 | 2021-10-08 | 江苏智通交通科技有限公司 | Abnormal driving early warning method based on reasonable driving range of vehicles at intersection |
| CN110728842A (en) * | 2019-10-23 | 2020-01-24 | 江苏智通交通科技有限公司 | Abnormal driving early warning method based on reasonable driving range of vehicles at intersection |
| CN112164223A (en) * | 2020-02-27 | 2021-01-01 | 周国霞 | Intelligent traffic information processing method and device based on cloud platform |
| CN112164223B (en) * | 2020-02-27 | 2022-04-29 | 浙江恒隆智慧科技集团有限公司 | Intelligent traffic information processing method and device based on cloud platform |
| CN111597905A (en) * | 2020-04-20 | 2020-08-28 | 淮阴工学院 | Highway tunnel parking detection method based on video technology |
| CN112673406A (en) * | 2020-05-29 | 2021-04-16 | 华为技术有限公司 | Method and terminal device for identifying abnormal vehicle parameters in vehicle queue |
| WO2021237745A1 (en) * | 2020-05-29 | 2021-12-02 | 华为技术有限公司 | Method for identifying abnormal vehicle parameters in vehicle platoon and terminal device |
| CN113658423A (en) * | 2020-10-23 | 2021-11-16 | 秦胜君 | Vehicle track abnormity detection method based on circulating gate control unit |
| CN112700474A (en) * | 2020-12-31 | 2021-04-23 | 广东美的白色家电技术创新中心有限公司 | Collision detection method, device and computer-readable storage medium |
| CN112700474B (en) * | 2020-12-31 | 2024-08-20 | 广东美的白色家电技术创新中心有限公司 | Collision detection method, apparatus, and computer-readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107644206A (en) | A kind of road abnormal behaviour action detection device | |
| CN110738857B (en) | Vehicle violation evidence obtaining method, device and equipment | |
| TWI452540B (en) | Image based detecting system and method for traffic parameters and computer program product thereof | |
| KR100459476B1 (en) | Apparatus and method for queue length of vehicle to measure | |
| CN110619279B (en) | Road traffic sign instance segmentation method based on tracking | |
| KR101908611B1 (en) | Parking control method for vehicle | |
| CN111753797B (en) | Vehicle speed measuring method based on video analysis | |
| CN108229256B (en) | Road construction detection method and device | |
| KR102282800B1 (en) | Method for trackig multi target employing ridar and camera | |
| CN109344886B (en) | Occlusion number plate distinguishing method based on convolutional neural network | |
| Liu et al. | Vision-based real-time lane marking detection and tracking | |
| CN106156695B (en) | Outlet and/or entrance area recognition methods and device | |
| CN111524350B (en) | Method, system, terminal device and medium for detecting abnormal driving condition of vehicle and road cooperation | |
| CN110781834A (en) | Traffic abnormality image detection method, device, computer device and storage medium | |
| Garg et al. | Real-time road traffic density estimation using block variance | |
| CN116109986A (en) | Vehicle track extraction method based on laser radar and video technology complementation | |
| CN114360261B (en) | Vehicle reverse running identification method and device, big data analysis platform and medium | |
| CN109344829B (en) | Train number identification method and device for high-speed railway train | |
| KR102494953B1 (en) | On-device real-time traffic signal control system based on deep learning | |
| Dinh et al. | Development of a tracking-based system for automated traffic data collection for roundabouts | |
| CN113537170A (en) | Intelligent traffic road condition monitoring method and computer readable storage medium | |
| CN117115752A (en) | Expressway video monitoring method and system | |
| Špaňhel et al. | Detection of traffic violations of road users based on convolutional neural networks | |
| Parida et al. | Recognition Of Vehicle Number Plate Using MATLAB | |
| KR20220089217A (en) | Two-wheeled vehicle control apparatus and method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180130 |
|
| RJ01 | Rejection of invention patent application after publication |