CN105809954A - Traffic event detection method and system - Google Patents
Traffic event detection method and system Download PDFInfo
- Publication number
- CN105809954A CN105809954A CN201410840963.2A CN201410840963A CN105809954A CN 105809954 A CN105809954 A CN 105809954A CN 201410840963 A CN201410840963 A CN 201410840963A CN 105809954 A CN105809954 A CN 105809954A
- Authority
- CN
- China
- Prior art keywords
- state
- characteristic point
- entropy
- interval
- traffic
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Traffic Control Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a traffic event detection method and system. The traffic event detection method comprises steps of a motion state quantity acquisition step, an entropy value calculation step, and a traffic event detection step. The motion state quantity acquisition step comprises a step of obtaining the motion state quantity of the motion variation of each time characteristic point, wherein the characteristic point expresses the vehicle; the entropy calculation step comprises a step of calculating the entropy value of the vehicle flow at each moment based on the obtained motion state quantity value of each moment; and the traffic event detection step comprises the steps of calculating the entropy of the vehicle flow at each moment to perform fitting per every predetermined time and determining whether the traffic accident happens according to the fitting result.
Description
Technical field
The present invention relates to traffic incidents detection method and system.
Background technology
In the past, in traffic incidents detection, realize the identification of traffic events often through complicated process.Such as traffic events video identification, in this process, need combining image process, pattern recognition, machine learning scheduling theory and method, such as there is techniques below: express and carry out further investigated first against the initialization background of background model, background with renewal, it is achieved that the self adaptation foreground moving region detection under complex scene;Morphological character and kinetic characteristic in conjunction with moving target devise the sorting algorithm based on multi-class support vector machine, the classification achieving mixed traffic moving target judges, propose the multiple features Matching pursuitalgorithm based on Kalman filtering and the occlusion handling method based on historical movement information compensation, it is ensured that the kinestate under complicated circumstance of occlusion is accurately estimated;Consider the spatial character of track, directional characteristic, category Properties, it is proposed that the multi-level learning method of motor pattern, and thus construct based on Bayes spatial model coupling and the anomaly detection method based on origin and destination direction mode coupling;The status attribute of moving target is combined with the context-related information in traffic scene, defines the concrete concept such as simple event, complicated event, provide general expression-form for event recognition;On this basis, the elementary event recognition methods combined based on Bayes grader and the complicated event recognition methods based on hidden Markov model are constructed with logical constraint.
It addition, in the past, carrying out in the method detected for Video Events, it is proposed that based on semantic Video Events determination method.The method utilizes a kind of general semantic meaning representation form to carry out traffic events reasonably describing and expressing, and then Land use models is known method for distinguishing and realized effectively automatically identifying of event.The method around video do more physical exercises object identification features describe with classification, based on semantic video complicated event detection with analyze, Context event relatedness is excavated and event level high level semantic description and understand the key technology of four aspects, it is proposed that the characteristics of objects of doing more physical exercises of adaptive combined invariant moment value describe with sorting technique, the detection of track multi-tag hypergraph model and analyze complicated event method, temporal sequence association rule excavates event semantics algorithm and case grammar network structure framework describes and understands video multithreading event technology.
Summary of the invention
But, carry out, in the apparatus and method detected, all there is various problem at above for traffic events.
Firstly, there are the problem that cost is high.Movement of traffic target at real world is diversified, in discriminator process, it is necessary to selects the feature of magnanimity, and then realizes comparatively accurate, careful target category division.The traffic behavior complexity of moving target is various, and randomness is stronger, it is necessary to gathers the track sample of magnanimity, and then sets up rationally effective behavioral pattern learning model.Needs structure is extensive, cover eurypalynous Video Events data base, the definition of event type and the unification of standard is easy to the expansion of video content semantic analysis correlational study work.
Additionally, there are baroque problem.The context-related information close association effectively identified with traffic events of traffic events, on the basis that target behavior is analyzed, in addition it is also necessary to strengthen the INTELLIGENT IDENTIFICATION of context.Video file comprises abundant semantic information, it is necessary to describes and extracts many attribute character in video, and find out relatedness therein.The research of video semanteme description standard.After analysis video semanteme, in addition it is also necessary to comprehensive, the accurate and general research on standard describing video semanteme.
Further, there is the problem that application is limited.Under current video detection technology, under night scenes, sleety weather, identifying moving object and classify, the stability of system work is still not high, causes that its application is very limited.
Therefore, the present invention proposes can judge rapidly traffic events and traffic incidents detection technology easy to use.
The present invention provides a kind of traffic incidents detection method, including: movement-state obtaining step, obtain the movement-state of the motion change of each moment characteristic point, described Based on Feature Points vehicle;Entropy calculation procedure, the value based on the described movement-state in acquired each moment calculates the wagon flow entropy in each moment;And traffic incidents detection step, it is fitted every the scheduled time wagon flow to the calculating described entropy in each moment, determines whether that traffic events occurs according to fitting result.
According to above-mentioned traffic incidents detection method, in described entropy calculation procedure, multiple state is divided interval based on the value of movement-state, calculate the quantity of characteristic point in each state interval and the ratio of number of the whole characteristic points probability as each described state interval, based on entropy described in described probability calculation.
According to above-mentioned traffic incidents detection method, described characteristic point is described vehicle itself, obtains described movement-state by vehicle checker in described movement-state obtaining step.
According to above-mentioned traffic incidents detection method, described characteristic point is the angle point of vehicle described in video image, in described movement-state obtaining step, based on described angle point coordinate in not video image in the same time, obtain the movement-state that each moment reflects the motion change of each described angle point.
According to above-mentioned traffic incidents detection method, described movement-state is any one in the position coordinates of the direction of the speed of the size of the speed of characteristic point, characteristic point, the acceleration of characteristic point, characteristic point or multinomial combination.
According to above-mentioned traffic incidents detection method, described movement-state is the combination in the direction of the speed of the size of the speed of described characteristic point and described characteristic point, in described entropy calculation procedure, direction based on the size of speed of described characteristic point and the speed of described characteristic point carries out the division that state is interval, calculates characteristic point number in each state interval that the direction of the size of the speed by described characteristic point and the speed of described characteristic point divides relative to the ratio of all characteristic point numbers as the interval probability of each state.
According to above-mentioned traffic incidents detection method, when dividing state interval, each state of decile is interval.
According to above-mentioned traffic incidents detection method, when dividing state interval, each state of decile is not interval.
According to above-mentioned traffic incidents detection method, also include: Corner Detection step, each pixel from video image detects angle point;And angle point tracking step, utilize optical flow tracking that described angle point is moved into line trace, wherein, described characteristic point is the angle point of described vehicle.
According to the traffic incidents detection method that entropy is stated, in traffic incidents detection step, calculate alarm level according to the slope of the straight line of entropy linear fit, the moment that the sliding window of linear fit is started and terminated as the beginning and ending time reported to the police.
The present invention relates to a kind of traffic incident detecting system, including: movement-state acquisition unit, described movement-state acquisition unit acquisition each moment reflects the movement-state of the motion change of each characteristic point, described Based on Feature Points vehicle;Entropy calculating part, described entropy calculating part calculates the wagon flow entropy in each moment based on the value of the described movement-state in each acquired moment;And traffic incidents detection portion, described traffic incidents detection portion is fitted every the scheduled time wagon flow to the calculating described entropy in each moment, determines whether that traffic events occurs according to fitting result.
According to above-mentioned traffic incident detecting system, it is interval that described entropy calculating part divides multiple state based on the value of movement-state, calculate the quantity of characteristic point in each state interval and the ratio of number of the whole characteristic points probability as each described state interval, based on entropy described in described probability calculation.
According to above-mentioned traffic incident detecting system, described characteristic point is described vehicle itself, obtains described movement-state by vehicle checker in described movement-state obtaining step.
According to above-mentioned traffic incident detecting system, described characteristic point is the angle point of vehicle described in video image, described movement-state acquisition unit, based on described angle point coordinate in not video image in the same time, obtains the movement-state that current time reflects the motion change of each described angle point.
According to above-mentioned traffic incident detecting system, described movement-state is any one in the position coordinates of the direction of the speed of the size of the speed of characteristic point, characteristic point, the acceleration of characteristic point, characteristic point or multinomial combination.
According to above-mentioned traffic incident detecting system, described movement-state is the combination in the direction of the speed of the size of the speed of described characteristic point and described characteristic point, described movement-state is the combination in the direction of the speed of the size of the speed of described characteristic point and described characteristic point, described entropy calculating part carries out, based on the direction of the size of the speed of described characteristic point and the speed of described characteristic point, the division that state is interval, calculate characteristic point number in each state interval that the direction of the size of speed by described characteristic point and the speed of described characteristic point divides relative to the ratio of all characteristic point numbers as the interval probability of each state.
According to above-mentioned traffic incident detecting system, when dividing state interval, each state of decile is interval.
According to above-mentioned traffic incident detecting system, when dividing state interval, each state of decile is not interval.
According to above-mentioned traffic incident detecting system, also including: Corner Detection portion, described Corner Detection portion detects angle point in each pixel from video image;And angle point tracking portion, described angle point tracking portion utilizes optical flow tracking that described angle point is moved into line trace, and wherein, described characteristic point is the angle point of described vehicle.
According to above-mentioned traffic incident detecting system, described traffic incidents detection portion calculates alarm level according to the slope of the straight line of entropy linear fit, the moment that the sliding window of linear fit is started and terminated as the beginning and ending time reported to the police.
According to the present invention, the video features of magnanimity need not be selected, accurate target classification and identification need not be realized, the track sample of magnanimity need not be gathered, the Video Events storehouse of complexity need not be built, the contextual information of traffic events need not be carried out INTELLIGENT IDENTIFICATION, it is not necessary to build abundant semantic base.Utilize entropy to reduce the complexity of system to detect traffic events, and and then reduce application cost.It addition, in accordance with the invention it is possible to detect event (i.e. traffic events) macroscopically by the micro state (i.e. entropy) of system.
It addition, in the traffic incidents detection utilizing video image, owing to adopting angle point, therefore angle point is little affected by the impact of illumination condition, under the weather such as night scenes, sleet, it is also possible to detect traffic events preferably.
Accompanying drawing explanation
Fig. 1 is the flow chart illustrating traffic incidents detection method;
Fig. 2 is illustrated based on the figure that state interval is divided by the size and Orientation of speed;
Fig. 3 is an illustration entropy time history plot;
Fig. 4 is the figure that the illustration curve to entropy is fitted;
Fig. 5 is the curve chart of an illustration alarm level and time of fire alarming;
Fig. 6 is the flow chart of the concrete mode illustrating traffic incidents detection method;
Fig. 7 is the structure chart illustrating traffic incident detecting system;
Fig. 8 is the structure chart of the concrete structure illustrating traffic incident detecting system;
Fig. 9 A~Fig. 9 C illustrates that figure, Fig. 9 A that vehicle congestion situation is detected illustrates that figure, Fig. 9 B of detected angle point is that figure, Fig. 9 C of the change illustrating entropy illustrates the time dependent figure of alarm level.
Detailed description of the invention
Below, with reference to accompanying drawing, the present invention is described in detail.
Fig. 1 illustrates the flow chart of traffic incidents detection method.First, the movement-state of the motion change of each moment characteristic point, wherein said Based on Feature Points vehicle are obtained;Then, the wagon flow entropy in each moment is calculated based on the value of the described movement-state in acquired each moment;Finally, it is fitted calculating the wagon flow described entropy in each moment every the scheduled time, determines whether that traffic events occurs according to fitting result.Traffic events mentioned here is the anomalous event instigating traffic that large change occurs at short notice.Traffic events has the feature being clearly distinguishable from normal traffic wagon flow, for instance vehicle accident, wagon flow sudden change etc..Specifically, traffic events be such as traffic congestion, collide, thrust into, vehicle abnormality travel speed, vehicle doubling, car hits railing, car hits roadbed etc..The sudden change of entropy, it is meant that traffic behavior there occurs sudden change, thus meaning to there occurs once new traffic events.
First, as shown in the step S1 of Fig. 1, obtaining the movement-state of vehicle, described movement-state can be the amounts relevant with speed such as the size of speed, the direction of speed, acceleration, position coordinates or the combination between them.The mode of the movement-state of acquisition vehicle can be by the internet of things equipment terminals such as vehicle checker or other approach obtain.But as long as movement-state can be obtained, it is not particularly limited.
As shown in the step S2 of Fig. 1, after obtaining movement-state, calculate entropy based on acquired movement-state.
It is velocity field for movement-state, calculates people in video, the entropy of velocity field that car collectively forms.Fig. 2 is illustrated based on the figure that state interval is divided by the size and Orientation of speed.As illustrated in fig. 2, it is assumed that total number of the vehicle followed the tracks of is N (namely total number of particles is N), according to the velocity attitude of each particle of current time and size, calculate the entropy in this moment.Calculation process is as follows:
The velocity attitude of particle is carried out interval division, from 0-360 degree, is divided into M decile.The velocity magnitude of particle is carried out interval division, is divided into the by stages such as N.
Calculate total state interval C (M, N)=M*N.
(i j), namely falls into state interval Q to the two-dimentional frequency H of calculating speedi*jParticle number, obtain and fall into the particle number of the by stages such as each state, i is greater than being equal to 1 integer less than or equal to M, and j is greater than being equal to 1 integer less than or equal to N.
(i, j), as in figure 2 it is shown, abscissa is velocity magnitude, vertical coordinate is velocity attitude to draw X-Y scheme H.
As shown in formula (1), calculate the probability P that each state is intervalij。
As shown in formula (2), calculate the entropy S in each moment
Entropy is the result sum that the interval probability of each state is multiplied by the log function of probability.
Here movement-state is not limited to the speed of vehicle, as long as the quantity of state of the motion of vehicle can be reflected.Such as movement-state can be only the size of the speed of vehicle, it is also possible to be only the direction of the speed of vehicle, it is also possible to be their combination.It addition, movement-state can also be the acceleration of vehicle, it is also possible to be the position coordinates of vehicle.
It addition, when dividing state interval based on movement-state, it is possible to decile state is interval, it is also possible to decile state is not interval.
By repeating each above-mentioned step, it is possible to calculate the entropy in all moment, and draw the time dependent curve of entropy.Fig. 3 illustrates an example of entropy time history plot.The abscissa of Fig. 3 is the time, and vertical coordinate is entropy.As it is shown on figure 3, low from the 1 to 10th moment entropy, system is in order state;Increasing from 11 to 20 moment entropy, system is in disordered state.
As shown in the step S3 of Fig. 1, entropy calculated in above-mentioned steps is utilized to carry out traffic incidents detection.Change according to entropy, detects the sudden change of traffic events and wagon flow.The sudden change of entropy, it is meant that traffic behavior there occurs sudden change, thus meaning to there occurs once new traffic events.
Entropy for each moment is fitted.Fig. 4 illustrates the example that entropy carries out linear fit.Here illustrate for linear fit.Entropy to all moment calculated, (assumes 2 seconds) at regular intervals and entropy is carried out linear fit.As shown in Figure 4, the 9 sublinear matchings of line segment as shown in numeral 1-9 have altogether been carried out.
The slope of every section of line segment according to linear fit illustrated above, calculates alarm level, and alarm level is that the slope of fit line is divided by 90 degree.The beginning and ending time reported to the police is the moment that sliding window starts and terminates.As it is shown in figure 5, abscissa is time of fire alarming, vertical coordinate is alarm level.The slope of the linear fit according to entropy, calculates the alarm level in each moment.
In the above-mentioned process obtaining alarm level based on entropy, however it is not limited to entropy is carried out curve fitting, as long as alarm level can be obtained, it is possible to be the matching of any other mode.
Above embodiments of the present invention are illustrated, it is possible to apply the present invention in the traffic incidents detection of video image.Concrete mode is as described below.
Illustrate for the situation being detected traffic events by video image.Here highlight and utilize angle point to obtain movement-state and then the entropy calculating carrying out in video image.
In the picture, angle point is the point of curvature maximum on the violent point of two dimensional image brightness flop or image border curve, it it is an important local feature, which determine the shape of key area in image, embody characteristic information important in image, the contour feature of master goal is had decisive action, has also just substantially grasped the shape of target once the contour feature that have found target, be of great significance so angle point has in target recognition, images match, image reconstruction.
Further, since angle point has rotational invariance, therefore angle point is little affected by the impact of illumination condition, and it is in 3 D scene rebuilding, estimation, target following, target recognition, image registration and the computer vision field such as mates and plays very important effect.
In video image, vehicle can be represented by the multiple angle points of self.The angle point of vehicle is such as the point that two contour lines intersect.To vehicle detection follow the trail of can be simplified to the multiple angle points to vehicle self detection follow the trail of realize.Based on the feature of above-mentioned angle point, in traffic events, the important local feature in video image being held by detected angle point, directly vehicle and personnel being identified thus avoiding.Vehicle is represented, it is possible in the simultaneously effective minimizing data volume retaining key character information so that during to image procossing, operand is greatly reduced by angle point.Therefore, as shown in Figure 6, first, Corner Detection can be carried out in step s 11 for video image, thus obtaining the angle point representing vehicle.
The method of Corner Detection is varied, but generally can be divided into 4 classes: the Corner Detection based on edge feature, the Corner Detection based on gray level image, the Corner Detection based on bianry image and the Corner Detection based on mathematical morphology.The principle of Corner Detection is construed to the most intuitively: in any two mutually orthogonal direction, has the point of large change.
Harris Corner Detection diagonal angle point can be adopted to detect, and harris angle point possesses rotational invariance.
As shown in the step S12 of Fig. 6, carry out optical flow tracking for the above-mentioned angle point being detected.
Light stream is divided into sparse optical flow and dense optical flow, each pixel of dense optical flow and speed or can be described as relevant with displacement, uses the method that dense optical flow is able to pursuit movement to have Horn-Shrunk method, also has block matching method.In the present invention to adopt sparse optical flow to illustrate.Need when calculating to specify one group of point (angle point) before tracked, then utilize pyramid LK optical flow algorithm, to moving into line trace.
As shown in the step S13 of Fig. 6, obtain the movement-state of each angle point.In the frame video of front and back two, if the time is t and t+1, the coordinate of angle point i respectively pit(x, y) and pi(t+1)(x, y).Coordinate based on time and angle point calculates the movement-state about this angle point i.This movement-state is such as the speed v of angle pointi=pi(t+1)(x, y)-pit(x, y).Wherein, the velocity component in x direction is vix=pi(t+1)(x)-pitX (), the velocity component in y direction is viy=pi(t+1)(y)-pit(y)。
As shown in the step S14 of Fig. 6, after obtaining movement-state, calculate entropy based on acquired movement-state.It is velocity field for movement-state, calculates people in video, the entropy of velocity field that car collectively forms.Step S2 and S3 in step S14 and S15 and Fig. 1 of Fig. 6 compares, and outside replacing the vehicle in embodiment 1 other than with angle point, other are identical with embodiment 1, and explanation is therefore not repeated.
Fig. 7 illustrates the structure chart of traffic incident detecting system.Traffic incident detecting system 1 shown in Fig. 7 includes movement-state acquisition unit 11, entropy calculating part 12 and traffic incidents detection portion 13.
Movement-state acquisition unit 11 obtains the movement-state that each moment reflects the motion change of each characteristic point, described Based on Feature Points vehicle;Entropy calculating part 12 calculates the wagon flow entropy in each moment based on the value of the described movement-state in each acquired moment;And traffic incidents detection portion 13 is fitted every the scheduled time wagon flow to the calculating described entropy in each moment, determine whether that traffic events occurs according to fitting result.
Entropy calculating part 12 divides multiple state interval based on the value of movement-state, calculates the quantity of characteristic point in each state interval and the ratio of number of the whole characteristic points probability as described each state interval, based on entropy described in described probability calculation.
Described characteristic point can be described vehicle itself, obtains described movement-state by vehicle checker in described movement-state acquisition unit 11.Described characteristic point can also be the angle point of vehicle in video image, and movement-state portion 11 based on described angle point coordinate in not video image in the same time, obtains the movement-state that each moment reflects the motion change of each described angle point.
Movement-state can be any one in the position coordinates of the direction of the speed of the size of the speed of characteristic point, characteristic point, the acceleration of characteristic point, characteristic point or multinomial combination
When the combination in direction of speed that movement-state is the size of speed of characteristic point and characteristic point, described entropy calculating part 12 carries out, based on the direction of the size of the speed of described characteristic point and the speed of described characteristic point, the division that state is interval, and the two-dimentional frequency of the characteristic point number being calculated as in each state interval that the direction of the size of speed by described characteristic point and the speed of described characteristic point divides for all characteristic points.
It addition, when dividing state interval, it is possible to each state of decile is interval, it is also possible to each state of decile is not interval.
Described traffic incidents detection portion 25 calculates alarm level according to the slope of the straight line of entropy linear fit, the moment that the sliding window of linear fit is started and terminated as the beginning and ending time reported to the police.
What the above-mentioned traffic incident detecting system of the present invention was concrete as shown in Figure 8, can utilize in image the angle point of vehicle to carry out the calculating of the movement-state of vehicle for video image.
Traffic incident detecting system shown in Fig. 8 also includes: Corner Detection portion 21, and described Corner Detection portion detects angle point in each pixel from video image;And angle point tracking portion 22, described angle point tracking portion utilizes optical flow tracking that described angle point is moved into line trace, and wherein, described characteristic point is the angle point of described vehicle.The traffic incident detecting system 20 of Fig. 8 utilizes Corner Detection and follows the tracks of the movement-state obtaining angle point, for instance the coordinate of angle point and the moment of correspondence can be utilized to calculate the speed of angle point.After the movement-state obtaining angle point, carry out the same process of Fig. 7, entropy calculating part 24 and traffic incidents detection portion 25, with entropy calculating part and the traffic incidents detection portion 25 of traffic incident detecting system 10, there is same function, narration is not repeated herein.
Application examples 1
With reference to Fig. 9 A to Fig. 9 C, detection vehicle congestion situation is illustrated.Fig. 9 A is the figure illustrating detected angle point, and wherein, the circle in Fig. 9 A represents that angle point, Fig. 9 B be figure, Fig. 9 C of the change illustrating entropy are illustrate the time dependent figure of alarm level.
Traffic incidents detection method according to the present invention and system.Following application can be carried out.Detect traffic events in real time, the traffic events of early warning can be detected;Traffic events is carried out real-time early warning, according to the entropy trendgram calculated in real time, future transportation trend is made anticipation, the traffic events that will occur is carried out early warning;To traffic events Realtime Alerts, according to the entropy catastrophe calculated in real time, traffic events being reported to the police, alarm level is mutated into direct ratio with entropy, suddenlys change more big, and alarm level is more high.
According to above-mentioned, the present invention provides can judge rapidly traffic events and traffic incidents detection method easy to use and system.
The present invention is described above by reference to specific embodiments of the invention, but those skilled in the art all understand, these specific embodiments can be carried out various amendment, combination and change, without departing from the spirit and scope of the present invention limited by claims or its equivalent.
Claims (18)
1. a traffic incidents detection method, including:
Movement-state obtaining step, obtains the movement-state of the motion change of each moment characteristic point, described Based on Feature Points vehicle;
Entropy calculation procedure, the value based on the described movement-state in acquired each moment calculates the wagon flow entropy in each moment;And
Traffic incidents detection step, is fitted every the scheduled time wagon flow to the calculating described entropy in each moment, determines whether that traffic events occurs according to fitting result.
2. traffic incidents detection method as claimed in claim 1, it is characterised in that
In described entropy calculation procedure, multiple state is divided interval based on the value of movement-state, calculate the quantity of characteristic point in each state interval and the ratio of number of the whole characteristic points probability as each described state interval, based on entropy described in described probability calculation.
3. traffic incidents detection method as claimed in claim 1 or 2, it is characterised in that
Described characteristic point is the angle point of vehicle described in video image,
In described movement-state obtaining step, based on described angle point coordinate in not video image in the same time, obtain the movement-state that each moment reflects the motion change of each described angle point.
4. traffic incidents detection method as claimed any one in claims 1 to 3, it is characterised in that
Described movement-state is any one in the position coordinates of the direction of the speed of the size of the speed of characteristic point, characteristic point, the acceleration of characteristic point, characteristic point or multinomial combination.
5. traffic incidents detection method as claimed any one in claims 1 to 3, it is characterised in that
Described movement-state is the combination in the direction of the speed of the size of the speed of described characteristic point and described characteristic point,
In described entropy calculation procedure, direction based on the size of speed of described characteristic point and the speed of described characteristic point carries out the division that state is interval, calculates characteristic point number in each state interval that the direction of the size of the speed by described characteristic point and the speed of described characteristic point divides relative to the ratio of all characteristic point numbers as the interval probability of each state.
6. traffic incidents detection method as claimed in claim 2, it is characterised in that
When dividing state interval, each state of decile is interval.
7. traffic incidents detection method as claimed in claim 2, it is characterised in that
When dividing state interval, each state of decile is not interval.
8. traffic incidents detection method as claimed any one in claims 1 to 3, it is characterised in that
Also include:
Corner Detection step, detects angle point in each pixel from video image;And
Angle point tracking step, utilizes optical flow tracking that described angle point is moved into line trace,
Wherein, described characteristic point is the angle point of described vehicle.
9. the traffic incidents detection method as according to any one of claim 1 to 8, it is characterised in that
In traffic incidents detection step, calculate alarm level according to the slope of the straight line of entropy linear fit, the moment that the sliding window of linear fit is started and terminated as the beginning and ending time reported to the police.
10. a traffic incident detecting system, including:
Movement-state acquisition unit, described movement-state acquisition unit acquisition each moment reflects the movement-state of the motion change of each characteristic point, described Based on Feature Points vehicle;
Entropy calculating part, described entropy calculating part calculates the wagon flow entropy in each moment based on the value of the described movement-state in each acquired moment;And
Traffic incidents detection portion, described traffic incidents detection portion is fitted every the scheduled time wagon flow to the calculating described entropy in each moment, determines whether that traffic events occurs according to fitting result.
11. traffic incident detecting system as claimed in claim 10, it is characterised in that
It is interval that described entropy calculating part divides multiple state based on the value of movement-state, calculates the quantity of characteristic point in each state interval and the ratio of number of the whole characteristic points probability as each described state interval, based on entropy described in described probability calculation.
12. the traffic incident detecting system as described in claim 10 or 11, it is characterised in that
Described characteristic point is the angle point of vehicle described in video image,
Described movement-state acquisition unit based on described angle point coordinate in not video image in the same time, obtains the movement-state that each moment reflects the motion change of each described angle point.
13. the traffic incident detecting system as according to any one of claim 10 to 12, it is characterised in that
Described movement-state is any one in the position coordinates of the direction of the speed of the size of the speed of characteristic point, characteristic point, the acceleration of characteristic point, characteristic point or multinomial combination.
14. the traffic incident detecting system as according to any one of claim 10 to 13, it is characterised in that
Described movement-state is the combination in the direction of the speed of the size of the speed of described characteristic point and described characteristic point, described entropy calculating part carries out, based on the direction of the size of the speed of described characteristic point and the speed of described characteristic point, the division that state is interval, calculates characteristic point number in each state interval that the direction of the size of the speed by described characteristic point and the speed of described characteristic point divides relative to the ratio of all characteristic point numbers as the interval probability of each state.
15. traffic incident detecting system as claimed in claim 11, it is characterised in that
When dividing state interval, each state of decile is interval.
16. traffic incident detecting system as claimed in claim 11, it is characterised in that
When dividing state interval, each state of decile is not interval.
17. traffic incident detecting system as claimed in claim 10, it is characterised in that
Also include:
Corner Detection portion, described Corner Detection portion detects angle point in each pixel from video image;And
Angle point tracking portion, described angle point tracking portion utilizes optical flow tracking that described angle point is moved into line trace,
Wherein, described characteristic point is the angle point of described vehicle.
18. the traffic incident detecting system as according to any one of claim 10 to 17, it is characterised in that
Described traffic incidents detection portion calculates alarm level according to the slope of the straight line of entropy linear fit, the moment that the sliding window of linear fit is started and terminated as the beginning and ending time reported to the police.
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410840963.2A CN105809954B (en) | 2014-12-30 | 2014-12-30 | Traffic incidents detection method and system |
SG11201705390RA SG11201705390RA (en) | 2014-12-30 | 2015-12-29 | Traffic event detection method and system |
PCT/CN2015/099526 WO2016107561A1 (en) | 2014-12-30 | 2015-12-29 | Traffic event detection method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410840963.2A CN105809954B (en) | 2014-12-30 | 2014-12-30 | Traffic incidents detection method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105809954A true CN105809954A (en) | 2016-07-27 |
CN105809954B CN105809954B (en) | 2018-03-16 |
Family
ID=56284291
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410840963.2A Active CN105809954B (en) | 2014-12-30 | 2014-12-30 | Traffic incidents detection method and system |
Country Status (3)
Country | Link |
---|---|
CN (1) | CN105809954B (en) |
SG (1) | SG11201705390RA (en) |
WO (1) | WO2016107561A1 (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108256380A (en) * | 2016-12-28 | 2018-07-06 | 南宁市浩发科技有限公司 | Road traffic exception automatic testing method |
CN108345894A (en) * | 2017-01-22 | 2018-07-31 | 北京同方软件股份有限公司 | A kind of traffic incidents detection method based on deep learning and entropy model |
CN109145732A (en) * | 2018-07-17 | 2019-01-04 | 东南大学 | A kind of black smoke vehicle detection method based on Gabor projection |
CN109887276A (en) * | 2019-01-30 | 2019-06-14 | 北京同方软件股份有限公司 | The night traffic congestion detection method merged based on foreground extraction with deep learning |
CN112053563A (en) * | 2020-09-16 | 2020-12-08 | 北京百度网讯科技有限公司 | Event detection method, device, equipment and storage medium for cloud control platform |
CN113255405A (en) * | 2020-02-12 | 2021-08-13 | 广州汽车集团股份有限公司 | Parking space line identification method and system, parking space line identification device and storage medium |
CN113286194A (en) * | 2020-02-20 | 2021-08-20 | 北京三星通信技术研究有限公司 | Video processing method and device, electronic equipment and readable storage medium |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112464734B (en) * | 2020-11-04 | 2023-09-15 | 昆明理工大学 | Automatic identification method for walking motion characteristics of quadruped based on vision |
CN113920728B (en) * | 2021-10-11 | 2022-08-12 | 南京微达电子科技有限公司 | Detection and early warning method and system for obstacles thrown on highway |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2007026301A (en) * | 2005-07-20 | 2007-02-01 | Matsushita Electric Ind Co Ltd | Stopping/low-speed vehicle detector and stopping/low-speed vehicle detection method |
CN101639983A (en) * | 2009-08-21 | 2010-02-03 | 任雪梅 | Multilane traffic volume detection method based on image information entropy |
CN102436740A (en) * | 2011-09-29 | 2012-05-02 | 东南大学 | Automatic detection method of traffic incident on highway |
CN102542797A (en) * | 2010-12-09 | 2012-07-04 | 财团法人工业技术研究院 | Image-based traffic parameter detection system and method and computer program product |
CN103971521A (en) * | 2014-05-19 | 2014-08-06 | 清华大学 | Method and device for detecting road traffic abnormal events in real time |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2816919B2 (en) * | 1992-11-05 | 1998-10-27 | 松下電器産業株式会社 | Spatial average speed and traffic volume estimation method, point traffic signal control method, traffic volume estimation / traffic signal controller control device |
CN100573618C (en) * | 2008-07-07 | 2009-12-23 | 山东省计算中心 | A kind of traffic intersection four-phase vehicle flow detection method |
-
2014
- 2014-12-30 CN CN201410840963.2A patent/CN105809954B/en active Active
-
2015
- 2015-12-29 WO PCT/CN2015/099526 patent/WO2016107561A1/en active Application Filing
- 2015-12-29 SG SG11201705390RA patent/SG11201705390RA/en unknown
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2007026301A (en) * | 2005-07-20 | 2007-02-01 | Matsushita Electric Ind Co Ltd | Stopping/low-speed vehicle detector and stopping/low-speed vehicle detection method |
CN101639983A (en) * | 2009-08-21 | 2010-02-03 | 任雪梅 | Multilane traffic volume detection method based on image information entropy |
CN102542797A (en) * | 2010-12-09 | 2012-07-04 | 财团法人工业技术研究院 | Image-based traffic parameter detection system and method and computer program product |
CN102436740A (en) * | 2011-09-29 | 2012-05-02 | 东南大学 | Automatic detection method of traffic incident on highway |
CN103971521A (en) * | 2014-05-19 | 2014-08-06 | 清华大学 | Method and device for detecting road traffic abnormal events in real time |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108256380A (en) * | 2016-12-28 | 2018-07-06 | 南宁市浩发科技有限公司 | Road traffic exception automatic testing method |
CN108345894A (en) * | 2017-01-22 | 2018-07-31 | 北京同方软件股份有限公司 | A kind of traffic incidents detection method based on deep learning and entropy model |
CN109145732A (en) * | 2018-07-17 | 2019-01-04 | 东南大学 | A kind of black smoke vehicle detection method based on Gabor projection |
CN109887276A (en) * | 2019-01-30 | 2019-06-14 | 北京同方软件股份有限公司 | The night traffic congestion detection method merged based on foreground extraction with deep learning |
CN109887276B (en) * | 2019-01-30 | 2020-11-03 | 北京同方软件有限公司 | Night traffic jam detection method based on fusion of foreground extraction and deep learning |
CN113255405A (en) * | 2020-02-12 | 2021-08-13 | 广州汽车集团股份有限公司 | Parking space line identification method and system, parking space line identification device and storage medium |
CN113286194A (en) * | 2020-02-20 | 2021-08-20 | 北京三星通信技术研究有限公司 | Video processing method and device, electronic equipment and readable storage medium |
CN112053563A (en) * | 2020-09-16 | 2020-12-08 | 北京百度网讯科技有限公司 | Event detection method, device, equipment and storage medium for cloud control platform |
Also Published As
Publication number | Publication date |
---|---|
SG11201705390RA (en) | 2017-08-30 |
CN105809954B (en) | 2018-03-16 |
WO2016107561A1 (en) | 2016-07-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105809954A (en) | Traffic event detection method and system | |
Atev et al. | A vision-based approach to collision prediction at traffic intersections | |
CN105513349B (en) | Mountainous area highway vehicular events detection method based on double-visual angle study | |
Liu et al. | A survey of vision-based vehicle detection and tracking techniques in ITS | |
CN104657740A (en) | Method and apparatus for segmenting an occupancy grid for a surroundings model of a driver assistance system for a vehicle | |
US20090319560A1 (en) | System and method for multi-agent event detection and recognition | |
JP2007128513A (en) | Scene analysis | |
CN101727672A (en) | Method for detecting, tracking and identifying object abandoning/stealing event | |
Piciarelli et al. | Surveillance-oriented event detection in video streams | |
CN102063613A (en) | People counting method and device based on head recognition | |
Aradhya | Object detection and tracking using deep learning and artificial intelligence for video surveillance applications | |
Malhi et al. | Vision based intelligent traffic management system | |
CN105654516A (en) | Method for detecting small moving object on ground on basis of satellite image with target significance | |
KR20140132140A (en) | Method and apparatus for video surveillance based on detecting abnormal behavior using extraction of trajectories from crowd in images | |
Rishika et al. | Real-time vehicle detection and tracking using YOLO-based deep sort model: A computer vision application for traffic surveillance | |
Lee et al. | Estimation and analysis of urban traffic flow | |
Khan | Estimating Speeds and Directions of Pedestrians in Real-Time Videos: A solution to Road-Safety Problem. | |
CN110889347B (en) | Density traffic flow counting method and system based on space-time counting characteristics | |
Bhaskar et al. | Enhanced and effective parallel optical flow method for vehicle detection and tracking | |
Liu et al. | A review of traffic visual tracking technology | |
Gunawan et al. | Geometric deep particle filter for motorcycle tracking: development of intelligent traffic system in Jakarta | |
Bernaś | Objects detection and tracking in highly congested traffic using compressed video sequences | |
Choi et al. | Event classification for vehicle navigation system by regional optical flow analysis | |
Sun et al. | Intelligent traffic accident detection system using surveillance video | |
Gu et al. | Real-Time Vehicle Passenger Detection Through Deep Learning |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |