CN112232257B - Traffic abnormality determination method, device, equipment and medium - Google Patents
Traffic abnormality determination method, device, equipment and medium Download PDFInfo
- Publication number
- CN112232257B CN112232257B CN202011158286.8A CN202011158286A CN112232257B CN 112232257 B CN112232257 B CN 112232257B CN 202011158286 A CN202011158286 A CN 202011158286A CN 112232257 B CN112232257 B CN 112232257B
- Authority
- CN
- China
- Prior art keywords
- position information
- image
- current frame
- determining
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
- G06V20/54—Surveillance or monitoring of activities, e.g. for recognising suspicious objects of traffic, e.g. cars on the road, trains or boats
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
The invention discloses a traffic abnormality determining method, a device, equipment and a medium, which are used for solving the problem of inaccurate traffic abnormality determination in the prior art. According to the embodiment of the invention, the first position information of each first object in the current frame and the second position information of each second object in the image frames of the preset number before the current frame image are used for determining the second position information of the corresponding second object in the image frames of the preset number of each first object in the current frame image, and whether traffic at the current moment is abnormal or not is determined according to the position information of the same object in the current frame image and any previous image frame through the traffic abnormality identification model, so that whether traffic abnormality occurs in a preset time period is determined, and the accuracy of traffic abnormality determination can be accurately improved.
Description
Technical Field
The present invention relates to the field of intelligent monitoring technologies and the field of image processing technologies, and in particular, to a method, an apparatus, a device, and a medium for determining traffic anomalies.
Background
With the rapid increase of the scale of China cities, the demands of urban traffic objects and pedestrians for traveling continuously exceed the increase of infrastructures such as urban roads, and sudden traffic anomalies at any position in the urban roads can cause urban traffic disturbance and spread to surrounding intersections or areas.
In the prior art, the method for determining traffic abnormality comprises the following steps:
1. a large amount of traffic video monitoring is established, manual inspection of key areas is carried out by command center personnel of the teams and the teams, and even traffic polices of the teams are arranged to carry out real-time inspection on the road surfaces and dredge traffic. The measures effectively improve the urban traffic efficiency to a certain extent, but the labor investment is large, the working efficiency of personnel is low, the problem discovery is slow, and the certainty is not necessarily accurate.
2. According to different abnormal classifications, different business logics are combined, however, the method is a good method for simple problems which can be clearly illustrated by mathematical models, but for complex and changeable traffic systems, model parameter adjustment problems which are difficult to overcome are encountered, and whether the traffic is abnormal cannot be accurately determined.
From this, it is clear that the conventional technology has a problem of inaccurate determination to some extent when determining traffic abnormality.
Disclosure of Invention
The invention provides a method, a device, equipment and a medium for determining traffic abnormality, which are used for solving the problem of inaccurate traffic abnormality determination in the prior art.
In a first aspect, an embodiment of the present invention provides a traffic anomaly determination method, where the method includes:
Acquiring first position information of each first object in the current frame image through a multi-target detection model which is completed through training;
determining second position information of each second object in the image frame according to the first position information of each first object and the second position information of the second object, wherein the second object corresponds to each first object of the current frame image in the image frame;
and determining whether the detection point is abnormal in traffic at the current frame moment based on the first position information of each first object in the current frame image and the second position information of the corresponding second object in the set number of image frames of the first object before the current frame image through the trained traffic abnormality recognition model.
In a first aspect, an embodiment of the present invention provides a traffic anomaly determination apparatus, including:
the multi-target detection model module is used for acquiring first position information of each first object in the current frame image through the trained multi-target detection model;
the determining module is used for determining second position information of each second object in the image frame according to the first position information of each first object and the second position information of the second object, and determining a second object corresponding to each first object of the current frame image in the image frame according to the set number of image frames before the current frame image;
The traffic abnormality recognition model module is used for determining whether the detection point is abnormal in traffic or not at the moment of the current frame through the trained traffic abnormality recognition model based on the first position information of each first object in the image of the current frame and the second position information of the second object corresponding to the first object in the set number of image frames before the image of the current frame.
In a third aspect, the present invention also provides an electronic device comprising at least a processor and a memory, the processor being configured to implement the steps of the traffic anomaly determination method as described in any one of the above when executing a computer program stored in the memory.
In a fourth aspect, the present invention also provides a computer-readable storage medium storing a computer program which, when executed by a processor, implements the steps of the traffic anomaly determination method as described in any one of the above.
According to the embodiment of the invention, the first position information of each first object in the current frame and the second position information of each second object in the image frames of the preset number before the current frame image are used for determining the second position information of the corresponding second object in the image frames of the preset number of each first object in the current frame image, and whether traffic at the current moment is abnormal or not is determined according to the position information of the same object in the current frame image and any previous image frame through the traffic abnormality identification model, so that whether traffic abnormality occurs in a preset time period is determined, and the accuracy of traffic abnormality determination can be accurately improved.
Drawings
In order to more clearly illustrate the technical solutions of the present invention, the drawings that are needed in the description of the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic diagram of a traffic anomaly determination process according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a mapping relationship between cosine values and angles according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a sample image according to an embodiment of the present invention;
FIG. 4 is an internal structure diagram of a traffic anomaly identification model provided by an embodiment of the present invention;
fig. 5 is a schematic structural diagram of a traffic abnormality determining apparatus according to an embodiment of the present invention;
fig. 6 is an electronic device provided in an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail below with reference to the accompanying drawings, it being apparent that the described embodiments are only some, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In order to improve accuracy of traffic abnormality determination, the embodiment of the invention provides a traffic abnormality determination method, device, equipment and medium.
Example 1:
fig. 1 is a schematic diagram of a traffic anomaly determination process according to an embodiment of the present invention, where the process includes the following steps:
s101: and acquiring first position information of each first object in the current frame image through the trained multi-target detection model.
The road congestion level determination provided by the embodiment of the invention is applied to electronic equipment, and the electronic equipment can be intelligent equipment such as image acquisition equipment, PC (personal computer) or a server.
In order to accurately identify the first position information of each first object in the current frame image, the embodiment of the invention inputs each frame image or the images of a set number of video frames at intervals in the video acquired by the image acquisition device into the multi-target detection model, and the first position information of each first object in the frame image is obtained.
The video acquired by the image acquisition equipment can be accessed to a video management center of a traffic police through protocols such as RTSP, GB28181 and ONVIF, and each frame of image is obtained by decoding the received video.
The multi-target detection model outputs category identification information of each object contained in an input image and position information of each object based on the input image, wherein the category identification information represents the type of the object, the position information can be the position of a detection frame corresponding to the object in the image, and specifically, the detection frame can be a rectangular frame and can be represented by coordinates of the upper left corner and the lower right corner of the rectangular frame. In order to ensure that the position of the object can be accurately determined, a coordinate system for determining the position reference can be predefined, and the coordinate system can be determined by taking the upper left corner of the image as an origin, taking the right direction as the positive x-axis direction, and taking the vertical downward direction as the positive y-axis direction.
Specifically, the recognition result of each object output by the multi-target to-be-measured model can be represented by a multi-object prediction array:
[image_id,ymin,xmin,ymax,xmax,score,class]
wherein mage_id represents a unique ID of an image input into the multi-target detection model; ymin, xmin, ymax, xmax, which is the position information of the object, score the confidence score of the object for the category, class, category ID to which the object belongs, e.g., category ID 01 for pedestrians, category ID 02 for bicycles, category ID 03 for private cars, category ID 04 for motorcycles, category ID 05 for aircraft … …
Because the embodiment of the invention is used for determining traffic abnormality, the position information of the object with the category ID of the preset category is acquired based on the detection result output by the multi-object detection model.
The multi-target detection model adopted by the embodiment of the invention adopts the EfficientDet detection model which takes the EfficientNet model as a backbone network, and the EfficientDet detection model uses fewer parameters and lower calculation cost, so that the achievement with highest precision can be obtained. Of course, when selecting the detection model, other neural network models can be adopted, specifically, the detection model can be selected according to own requirements, and the selected multi-target detection model is not limited in the embodiment of the invention.
S102: and determining second position information of each second object in the image frame according to the first position information of each first object and the second position information of the second object, wherein the second object corresponds to each first object of the current frame image in the image frame.
In order to accurately determine the second object corresponding to each first object in the current frame image in the set number of image frames before the current frame image, after the first position information of each first object in the current frame image is obtained, because the position information of each corresponding object in the set number of image frames before the current frame image has been obtained, the objects contained in the set number of image frames before the current frame image are referred to as second objects for distinction, the position information of the second objects is referred to as second position information, and the second object corresponding to each first object in the current frame image in the set number of image frames before the current frame image can be determined according to the first position information of each first object and the second position information of each second object.
Specifically, since the position information of the object in the set number of image frames before the current frame image can be determined, the trajectory of each object can be determined, and thus, according to the trajectory of each object and the position of each first object in the current frame image, it can be determined which second object the first object corresponds to, that is, the same object can be identified in different frame images.
S103: and determining whether the detection point is abnormal in traffic at the current frame moment based on the first position information of each first object in the current frame image and the second position information of the corresponding second object in the set number of image frames of the first object before the current frame image through the trained traffic abnormality recognition model.
After determining the second position information of the second object corresponding to each first object in the image of the current frame in the previous set number of image frames, in order to accurately determine whether the current frame time is abnormal, in the embodiment of the present invention, a vector is determined for the first position information of each first object in the current frame time and the second position information of the second object corresponding to the first object in any one of the previous set number of image frames. And determining whether the current frame moment is abnormal or not based on the vector by training the traffic abnormality recognition model.
According to the embodiment of the invention, the first position information of each first object in the current frame and the second position information of each second object in the image frames of the preset number before the current frame image are used for determining the second position information of the corresponding second object in the image frames of the preset number of each first object in the current frame image, and the one-field traffic recognition model is used for determining whether the object is abnormal or not according to the position information of the current frame image and the same object in any image frame before, so that whether traffic abnormality occurs in a preset time period or not is determined, and the accuracy of traffic abnormality determination can be accurately improved.
Example 2:
in order to accurately determine the second object corresponding to each first object of the current frame image in the image frames of the set number before the current frame image, in the embodiment of the present invention, the first objects include at least one of the following: pedestrians, bicycles, private cars, motorcycles, buses, and trucks.
In order to determine traffic anomalies based on the above categories, in the embodiment of the present invention, after determining the category and position information of each object in the current frame image based on the multi-target detection model, the objects of which category IDs are pedestrians, bicycles, private cars, motorcycles, buses, and trucks are acquired and are referred to as first objects for the purpose of distinguishing.
Example 3:
in order to accurately determine the second object corresponding to each first object in the current frame image in the image frames of the set number before the current frame image, in the embodiments of the present invention, determining the second object corresponding to each first object in the current frame image in the image frames according to the first position information of each first object and the second position information of the second object includes:
Determining the distance and the angle between the first position information of each first object and the second position information of the second object in the same category of the first object according to the first position information of each first object and the second position information of the second object;
for each first object, judging whether the angle between the first object and a second object of the same class with the smallest distance from the first object is smaller than a preset angle threshold value or not; if yes, determining the second object as a second object corresponding to the first object; if not, deleting the distance with the minimum distance, and re-determining whether the angle between the first object and the second object with the same category with the minimum distance with the first object is smaller than the preset angle threshold value.
In order to accurately determine a second object corresponding to a first object of a current frame image in a set number of image frames before the current frame image, the embodiments of the present invention firstly classify the objects according to the category to which the first object belongs, and determine the correspondence relationship of the object according to the category of the object.
Since the first position information of each first object in the current frame image and the second position information of each second object in the image frames set a number before the current frame image are known, the distance and angle between each first object and each second object in the same category can be determined.
When determining the distance and angle between each first object and each second object, the centroid of each first object and each second object can be determined according to the position information of each first object and each second object, and the distance and angle between each first object and each second object can be determined according to the distance and angle between the centroids.
For example, in the embodiment of the present invention, the first position information and the second position information may be determined by the coordinates of the upper left corner and the lower right corner, and a certain object is taken as an example for illustration, x min For the abscissa of the upper left corner of the object detection frame, x max For the abscissa of the lower right corner of the object detection frame, y min For the ordinate of the upper left corner of the object detection frame, y max For the ordinate of the lower right corner of the object detection frame, the coordinates of the centroid of the object are expressed as (x, y), where x= (x) min +x max )/2，y＝(y min +y max )/2. Since the position of the image capturing device is fixed, the actual position represented by each pixel point in the image captured by the image capturing device is determined.
After the distance and angle between each first object and each second object are obtained, since the distances between the second object corresponding to the first object and the first object are the smallest among all the second objects in the same category for the first object, based on the principle, after the distances between each first object and each second object in the same category are obtained, the second object with the smallest distance from the first object is selected among the distances between the first object and each second object for each first object.
Since the driving direction of the same object does not change too much in a fixed time, that is, the angle change is not too large, it is determined whether the angle between the first object and the corresponding second object is within the angle threshold range, and if the angle is within the angle threshold range, it is determined that the determined second object corresponding to the first object is correct. If not located at the cornerAnd if the second object corresponding to the first object is incorrect, the second object corresponding to the first object is redetermined, and the second object is not consulted, so that the accuracy of traffic abnormality determination is further ensured. For example, inIn the second time, if the vehicle is not turned over and does not drift, the speed of the object is not more than 60Km/h, the distance traveled is not more than 0.5 m, and the angle theta is not changed by more than 20 degrees. The specific setting of the angle threshold ω is not limited here.
Therefore, in the embodiment of the present invention, the first object and the second object of the same class are determined, when determining the second object corresponding to the first object, it is first determined whether the angle between the first object and the second object with the smallest distance is smaller than a preset angle threshold, if yes, it is indicated that the second object is the second object corresponding to the first object in the frame image, if not, it is indicated that the second object is not the second object corresponding to the first object in the current frame image, and the rest of the second objects are re-determined.
Specifically, after the centroid coordinates of each first object in the current frame image are determined, the centroid coordinates of each first object in the same category can be put into the same array t ：[[x 1 ,y 1 ]，[x 2 ,y 2 ]，[x 3 ,y 3 ]，…]，centroid t Representing a centroid array for the t-th image frame. According to the centroid array of the t-th image frame and the centroid array of the t-1 th image frame, the distance and the angle between the centroid of each object in the t-th image frame and the centroid of each object in the same category in the t-1 th image frame can be obtained, and the distance and the angle can be expressed in a matrix form, and specifically see table 1:
TABLE 1
The first row in table 1 represents the centroid coordinates of each object in the t-1 th image frame, the first column represents the centroid coordinates of each object in the t-1 th image frame, wherein m objects are in the t-1 th image frame, n objects are in the t-th image frame, and the intersection positions of the i+1th row and the j+1th column represent the distances between the centroid of the j-th object in the t-1 th image frame and the centroid of the i-th object in the t-th image frame, which can be specifically determined according to the following formula:
wherein x is (t，i) An abscissa representing the centroid of the ith object in the ith image frame, y (t，i) An ordinate representing the centroid of the ith object in the ith image frame, x (t-1，j) An abscissa representing the mass center of the jth object in the t-1 th image frame, y (t-1，j) Representing the ordinate of the centroid of the jth object in the t-1 image frame.
The difference vector direction angle Θ is determined by:
wherein x is (t，i) An abscissa representing the centroid of the ith object in the ith image frame, y (t，i) An ordinate representing the centroid of the ith object in the ith image frame, x (t-1，j) An abscissa representing the mass center of the jth object in the t-1 th image frame, y (t-1，j) Representing the ordinate of the centroid of the jth object in the t-1 image frame.
Fig. 2 is a schematic diagram of a correspondence relationship between cosine values and angles according to an embodiment of the present invention. As can be seen from fig. 2, if the horizontal axis in the image is taken as the angle zero point, the angles are calculated clockwise, the same cosine value corresponds to two angles in two quadrants, the two angles in the first quadrant and the fourth quadrant correspond to the same cosine value, the two angles in the second quadrant and the third quadrant correspond to the same cosine value, and the angles need to be calculated separately. And because the connecting line of the centroid coordinates of the first object and the origin of coordinates can form a first vector, and the connecting line of the centroid coordinates of the second object corresponding to the first object and the origin of coordinates can form a second vector, wherein the direction angle of the connecting line of the first object and the second object can be represented by a vector v obtained by subtracting the second vector from the first vector, the vector v has corresponding vector coordinates, the angle determined by the cosine value is corrected according to the y value of the ordinate of the vector corresponding to the vector v, if y is positive, the value of theta is unchanged, and if y is negative, the value of theta is adjusted to 360-theta, and the determined angle theta is restored.
In the embodiment of the present invention, the angles are uniformly divided into [0 °,180 ° ] according to the value of cos Θ, then the angles Θ are re-corrected according to the positive and negative of the y value of the vector v, for example, when the value of cos Θ is 1/2, the corresponding angles may be 60 ° or 300 °, firstly, the angles Θ are uniformly determined to be 60 ° according to cos Θ=1/2, then the vector ordinate y corresponding to the vector v is determined to be positive or negative, if positive, the angles Θ are determined to be 60 °, and if negative, the angles Θ are determined to be 360 ° -60 ° =300°.
The distance and angle between the centroid of each object in the t-1 image frame and the centroid of each object in the t-1 image frame will result in a matrix:
D(t,t-1)＝[
[d(t,1,t-1,1)]Θ(t,1,t-1,1),[d(t,1,t-1,2)]Θ(t,1,t-1,2),[d(t,1,t-1,2)]Θ(t,1,t-1,2),…,,
[d(t,2,t-1,1)]Θ,[d(t,2,t-1,2),)]Θ,[d(t,2,t-1,3),]Θ,…,,
…
]
wherein the ith row in the vector represents the distance and angle between the ith object in the ith image frame and each object in the t-1 th image frame, and wherein the jth column in the vector represents the distance and angle between the jth object in the t-1 th image frame and each object in the t-1 th image frame.
Specifically, when selecting an object in a t-1 image frame corresponding to an object in each t image frame based on the vectors, selecting a corresponding column with a minimum distance for each row of the obtained vectors, judging whether an angle corresponding to the column is smaller than a preset angle threshold, if so, obtaining the object of which t-1 image frame the object corresponds to; if not, selecting the corresponding column with the minimum distance value except the distance, and continuing to judge. For example, the distance in the obtained column vector is [9,3,5, … ], wherein 9,3,5 and the like are the column vectors corresponding to the exemplified row vector, and refer to the column vector in the first row, wherein the distance value to the 9 th element is the smallest, that is, when determining the object in the t-1 th image frame corresponding to the first object in the t-1 th image frame, the determination is performed according to the 9 th object in the t-1 th image frame; in the vectors of the second row, the distance value to the 3 rd element is smallest, that is to say, when determining the object in the t-1 th image frame corresponding to the second object in the t-1 th image frame, the determination is made according to the object in the t-1 th image frame and the 3 rd object.
If the system processes multiple paths of camera video frame images at the same time, each path of camera video frame image is calculated independently.
Example 4:
in order to accurately determine a second object corresponding to each first object in the current frame image in the previous set number of image frames, in the embodiments of the present invention, after determining a distance between the first position information of each first object and the second position information of the second object in the same category of the first object according to the first position information of each first object and the second position information of the second object, the method includes, before determining, for each first object, whether an angle between the first object and the second object in the same category with the smallest distance from the first object is smaller than a preset angle, whether the angle between the first object and the second object in the same category with the smallest distance from the first object is smaller than the preset angle:
if the distance between the first object and the second object is larger than the set distance threshold value, deleting the identified distance, and executing the subsequent operation of judging whether the angle between the first object and the second object of the same category with the smallest distance between the first object and the second object is smaller than the preset angle.
In the embodiment of the present invention, because the distance between the first object at the time corresponding to the current frame image and the corresponding second object at the time corresponding to the previous set number of image frames is not too large in the case of road speed limiting or the like, if the distance between the first object and the certain second object is recognized to be greater than the set distance threshold, the recognized distance is deleted, and when the second object corresponding to the first object is determined, the distance is not considered, that is, the second object cannot be the second object corresponding to the first object.
For example, a private car is driven at a speed of 150Km/h, images in a video are acquired at 25 frames/second, the moving distance of an object in a current frame image and a last detected frame image is less than (120×1000)/(3600×25) =1.3 m/frame, and since the images are not measured by distance values, the distance threshold can be determined by determining the relative size of the object and the private car by 5 m, 1.3 m corresponds to the length of 1/4 objects in the images, the distance threshold can be set according to the distance, and the distance exceeding the distance threshold can be filtered, that is, when a second object with the smallest distance to the first object is selected, the distance exceeding the distance threshold is not referred to, so that the second object corresponding to the first object in the last frame image can be determined more accurately. And the speeds corresponding to different categories are not consistent, so that the distance thresholds for judging corresponding to different categories are not the same.
In order to accurately determine the second object corresponding to each first object in the current frame image in the previous set number of image frames, on the basis of the above embodiments, in the embodiments of the present invention, the method further includes:
for each first object, judging whether a second object corresponding to the first object is identical to a second object corresponding to other first objects, and if so, re-determining the second object for the first object and other first objects with the same second object as the first object.
Since there may be a case where one of the objects exceeds the other objects, there may be a case where one of the first objects corresponds to a plurality of the second objects, that is, there may be a case where one of the second objects has a minimum distance from the plurality of the first objects, so if the second object corresponding to the first object is identical to the second object corresponding to the other first objects, the second object corresponding to the plurality of the first objects is filtered out, that is, when the second object having the minimum distance from the first object is selected, the second object is not referred to, and when the second object corresponding to the other first object is selected, the second object is not considered any more, and the selection is performed again among the remaining second objects.
Example 5:
in order to accurately determine whether traffic is abnormal, in the embodiments of the present invention, the second position information of each second object in the image frame is determined for a set number of image frames before the current frame image, where the method further includes:
determining a third object which is blocked in the current frame according to the track of each object determined in the image frames of the set number before the current frame;
and determining third position information of the third object in the current frame image according to the track of each object in the video detection frame before the current frame.
Since there may be an occluded third object in the current frame image, in order to more accurately determine whether the current traffic is abnormal, third location information of the occluded third object needs to be determined.
Because the speed at which the object travels is generally fixed, and the position of the camera that acquires each image is fixed, the acquisition range corresponding to each image of the camera is fixed, the duration of the object from entering the acquisition range to exiting the acquisition range is substantially fixed, and the number of image frames within the duration is also determined, the maximum duration of timeout of the object can be set based on the duration, and the timeout set by different classes of objects is inconsistent, and the reduction of each of the maximum durations is determined from the number of image frames that can be processed within the maximum duration. When a first object with no track exists before is identified in the image frame, determining that the first object is an object which just enters the acquisition range, setting the timeout duration of the object to be the maximum duration based on the first object, so that each object which enters the acquisition range corresponds to one timeout duration, reducing the timeout duration by a preset damping value when one frame of image is detected, and indicating that the object has exited the acquisition range when the timeout duration of the object is 0.
When the timeout duration of the object is not 0, if the object is not acquired in the current frame image, the object is blocked by other objects. Therefore, according to the stored timeout duration of each first object, the number of objects in the acquisition range can be determined, according to the determined number of first objects, the number of blocked objects can be determined, and according to the track of each object determined before the current frame and the second position information of the second object, the third position information of the blocked third object can be determined. Because the trajectory of the third object can be determined based on the previously set number of image frames, the position information of the third object can be predicted for the current frame.
And in order to ensure the accuracy of the third position information determination, for the centroid of tracking loss, detecting whether the predicted position has a frame of other objects, if so, indicating that the object is possibly blocked by other objects, and if not, indicating that the predicted third position information is inaccurate, adding the tracking track again, and determining the third position information again.
Example 6:
in order to accurately determine whether traffic is abnormal, based on the above embodiments, in the embodiment of the present invention, the traffic abnormality recognition model is trained by:
For each sample image in a sample training set, determining a vector corresponding to the sample image according to sample first position information of each sample first object in the sample image and sample second position information of a second object corresponding to the sample first object in a set number of sample images before the current sample image;
determining a corresponding output value based on the vector through a traffic anomaly identification model;
and adjusting parameters of the traffic abnormality recognition model according to the output value.
In order to realize training of the traffic anomaly recognition model, a sample set for training is stored in the embodiment of the invention, and sample images in the sample set are images of traffic normal acquired by different traffic road sections.
For facilitating training of the traffic anomaly recognition model, the sample set stores, for each sample image, position information and identification information of an object contained in the sample image, and the specific identification information may be used to indicate a type of the object contained in the sample image, for example, 01 is a pedestrian, 02 is a bicycle, 03 is a private car, 04 is a motorcycle, 05 is a bus, 06 is a truck, and the like. And the sample images of the same road section are also marked with the position information of the same object in different sample images. Fig. 3 is a schematic diagram of an image of one sample provided by the embodiment of the present invention, and fig. 3 includes different participating subjects of the traffic flow.
In the embodiment of the invention, in order to realize the training of the traffic abnormality recognition model, a vector is determined according to the position information of the same object in different image frames in the sample image:
x= (object type ID 1X t-m，1 X t-m+1,1 …X t+1,1
Object type ID 2X t-m，2 X t-m+1,2 …X t+1,2
……
Object type IDn X t-m，n X t-m+1,n …X t+1,n )
Wherein the first row represents the type of the first object and its position information from the t-m image frame to the t-th image frame, X t-m，1 Representing the position information of the first object at the t-m image frame. And n× (m+2) data points in the vector X, wherein n× (m+1) is another column of object type ID of the position information, if the objects in the image frame are less than n, the rest vector points are subjected to zero padding processing, if the objects in the image frame are more than n, the objects are screened, and n of the image frames are determined, wherein if the objects in the image frame are more than n, the object types are distinguished according to human experience firstly, and the traffic body comprises static characteristics anddynamic features, static features are generally not prone to traffic anomalies, so objects of the static features are first separated from image frames, thereby obtaining n objects. Specific static features correspond to different types of objects, such as highways, for different traffic segments, the static features include pedestrians, how the specific static features correspond to the traffic segments is not limited here, and how objects in the image frames are screened is not limited here.
And because the object type ID can influence the final output when passing through the traffic anomaly recognition model, in order to prevent local characteristics from greatly influencing network output fluctuation, preprocessing type normalization processing is performed on the ID according to the object motion condition under the acquired time interval, so that fluctuation of the traffic anomaly model caused by unbalance in data is avoided. Specifically, how to perform the pretreatment normalization is the prior art, and is not described herein.
Inputting the vector into a traffic abnormality recognition model to obtain an output value, wherein the output value is a value larger than 0, and if the output value is a value smaller than 0, the abnormal occurrence of traffic is indicated; and adjusting parameters of the traffic abnormality recognition model according to the determined output value.
Fig. 4 is an internal structure diagram of a traffic abnormality recognition model according to an embodiment of the present invention. This is described in detail by taking fig. 4 as an example.
Inputting a vector X into a first hidden layer of a traffic anomaly identification model, and linearly changing the vector X into G1 (X) to obtain G 1 (X)＝W 1 X+b 1
Wherein W is 1 And b 1 The linear characteristic of the vector is learned in the training process by the traffic anomaly identification model so as to obtain a first parameter and a first offset value, and each traffic data is equivalent to a segmentation curved surface, because the embodiment of the invention judges whether traffic anomaly occurs or not by the distance between the output data point and the segmentation curved surface, and if the segmentation curved surface exists, a plurality of available segmentation curved surfaces are infinite, the distance between the data point and the segmentation curved surface is inconvenient to calculate, therefore, after W1 and b1 are obtained by learning, normalization operation is respectively carried out on the data point and the segmentation curved surface, And the calculation of training loss and the rapid convergence of the model are facilitated. The normalization is performed as follows:
after normalization, G 'is obtained' 1 (x)＝W′ 1 X+b′ 1
Since urban traffic is a variety of conditions, and very complex, traffic data separation by a single separation plane will be subject to a very large number of false positives. Introduction of an activation function H (G' 1 (X)) for G' 1 The (X) output is subjected to nonlinear operation, G' 1 The separation curved surface mapped by (X) is not continuous, but a plurality of non-differentiable point sets exist, so that the adaptability of the separation curved surface to different abnormal situations is enhanced.
After the H function activation operation, a second hidden layer is introduced, and nonlinear characteristics of the vector are learned, so as to obtain a second parameter W 2 And a second bias b 2 To obtain a function G 2 (Y)＝W 2 Y+b 2 Wherein y=h (G' 1 (X))。
For the same reason as described above, it is necessary to normalize W2 and b2 in the following manner:
G′ 2 (Y)＝W′ 1 Y+b′ 1
through the hidden layer of the second layer, the traffic anomaly recognition model introduces a high-dimensional nonlinear separation hypersurface, and the capability of the model for coping with complex traffic scenes is further enhanced.
Traffic data passing through the traffic anomaly recognition model is finally divided into two parts, normal traffic and abnormal traffic. For example, the final output for a normal traffic input X would be O (G 2 (H(G′ 1 (X))) is more than or equal to 0, and all objects cannot fall into the inside of the separation curved surface due to the diversity and complexity of traffic anomalies, so elasticity can be introduced to further increase the capability of the traffic anomaly recognition model that different traffic anomalies should be obtained, and if zeta introduced by elasticity of the output result after consideration is taken into consideration i Then O (G) 2 (H(G′ 1 (X)))≥0-ξ i . Wherein xi i Is equal to the O (G) 2 (H(G′ 1 (X)) related values. In particular, the output value is not limited herein as to how to determine whether the traffic is abnormal.
In order to accurately determine whether an abnormality occurs in traffic, in the embodiments of the present invention, the adjusting parameters of the traffic abnormality recognition model according to the output value includes:
according toξ i ＝max(0,1-O(G 2 (H(G′ 1 (X))) adjusting parameters of the traffic anomaly identification model, wherein W 1 Is a first parameter, W 2 N is the number of vectors commonly input to the traffic anomaly recognition model, O (G 2 (H(G′ 1 (X)) is an output value corresponding to the vector, and γ is a parameter value set in advance.
In the embodiment of the invention, in order to accurately realize the training of the traffic abnormality recognition model, the traffic abnormality recognition model is used for training the traffic abnormality recognition model according to the following conditionsξ i ＝max(0,1-O(G 2 (H(G′ 1 (X)) for adjusting the parameters of the traffic abnormality recognition model.
Wherein L is a loss value determined by the model, and when the loss value is larger than a preset value And adjusting parameters in the traffic abnormality recognition model. And wherein W is 1 Is a first parameter, W 2 N is the number of vectors commonly input to the traffic anomaly recognition model, O (G 2 (H(G′ 1 (X)) is an output value corresponding to the vector, gamma is a parameter value set in advance,for representing xi i Weights in the loss function, and +.>Typically a value of (0, 1), gamma being a value greater than 1, ζ i An absolute value corresponding to an output value obtained by subtracting the ith vector from 0 or 1, if 0-O (G 2 (H(G′ 1 (X)))＞1-O(G 2 (H(G′ 1 (X))), then ζ i ＝|1-O(G 2 (H(G′ 1 (X))); if 0-O (G) 2 (H(G′ 1 (X)))＜1-O(G 2 (H(G′ 1 (X))), then ζ i ＝|0-O(G 2 (H(G′ 1 (X)))|。
Example 7:
fig. 5 is a schematic structural diagram of a traffic abnormality determining apparatus according to an embodiment of the present invention, where the apparatus includes:
the multi-target detection model module 501 is configured to acquire first position information of each first object in the current frame image through a trained multi-target detection model;
a determining module 502, configured to determine, for a set number of image frames before the current frame image, second position information of each second object in the image frame, and determine, according to the first position information of each first object and the second position information of the second object, a second object corresponding to each first object in the current frame image in the image frame;
The traffic anomaly recognition model module 503 is configured to determine whether the detection point is abnormal in traffic at the current frame time based on the first position information of each first object in the current frame image and the second position information of the corresponding second object in the set number of image frames of the first object before the current frame image through the trained traffic anomaly recognition model.
In a possible implementation manner, the determining module 502 is specifically configured to determine, according to the first position information of each first object and the second position information of the second object, a distance and an angle between the first position information of each first object and the second position information of the second object in the same category of the first object; for each first object, judging whether the angle between the first object and a second object of the same class with the smallest distance from the first object is smaller than a preset angle threshold value or not; if yes, determining the second object as a second object corresponding to the first object; if not, deleting the distance with the minimum distance, and re-determining whether the angle between the first object and the second object with the same category with the minimum distance with the first object is smaller than the preset angle threshold value.
In one possible embodiment, the apparatus further comprises: and the processing module is used for deleting the identified distance if the distance between the processing module and the first object is larger than the set distance threshold value, and executing the subsequent operation of judging whether the angle between the first object and the second object of the same type with the smallest distance between the processing module and the first object is smaller than the preset angle for each first object.
In one possible implementation manner, the processing module is further configured to determine, for each first object, whether a second object corresponding to the first object is the same as a second object corresponding to another first object, and if so, re-determine the second object for the first object and another first object having the same second object as the first object.
In a possible implementation manner, the determining module 502 is further configured to determine, according to the trajectory of each object determined in the set number of image frames before the current frame, a third object that is occluded in the current frame; and determining third position information of the third object in the current frame image according to the track of each object in the video detection frame before the current frame.
In one possible embodiment, the apparatus further comprises: the training module is used for determining a vector corresponding to each sample image in the sample training set according to the first position information of the sample of each sample first object in the sample image and the second position information of the sample of the corresponding second object in the set number of sample images before the current sample image; determining a corresponding output value based on the vector through a traffic anomaly identification model; and adjusting parameters of the traffic abnormality recognition model according to the output value.
In a possible implementation manner, the training module is further configured to, according to the following ξ i ＝max(0,1-O(G 2 (H(G′ 1 (X)))), adjusting parameters of the traffic anomaly identification model, wherein W 1 Is a first parameter, W 2 N is the number of vectors commonly input to the traffic anomaly recognition model, O (G 2 (H(G′ 1 (X)) is an output value corresponding to the vector, and r is a preset radius of the dividing plane.
Example 8:
on the basis of the above embodiments, the embodiment of the present invention further provides an electronic device, as shown in fig. 6, including: processor 601, communication interface 602, memory 603 and communication bus 604, wherein processor 601, communication interface 602, memory 603 accomplish each other's communication through communication bus 604.
The memory 603 has stored therein a computer program which, when executed by the processor 601, causes the processor 601 to perform the steps of:
acquiring first position information of each first object in the current frame image through a multi-target detection model which is completed through training;
determining second position information of each second object in the image frame according to the first position information of each first object and the second position information of the second object, wherein the second object corresponds to each first object of the current frame image in the image frame;
and determining whether the detection point is abnormal in traffic at the current frame moment based on the first position information of each first object in the current frame image and the second position information of the corresponding second object in the set number of image frames of the first object before the current frame image through the trained traffic abnormality recognition model.
In one possible embodiment, the first object comprises at least one of: pedestrians, bicycles, private cars, motorcycles, buses, and trucks.
In one possible implementation manner, the determining, according to the first position information of each first object and the second position information of the second object, the second object corresponding to each first object in the current frame image in the image frame includes:
Determining the distance and the angle between the first position information of each first object and the second position information of the second object in the same category of the first object according to the first position information of each first object and the second position information of the second object;
for each first object, judging whether the angle between the first object and a second object of the same class with the smallest distance from the first object is smaller than a preset angle threshold value or not; if yes, determining the second object as a second object corresponding to the first object; if not, deleting the distance with the minimum distance, and re-determining whether the angle between the first object and the second object with the same category with the minimum distance with the first object is smaller than the preset angle threshold value.
In one possible implementation manner, after determining, according to the first position information of each first object and the second position information of the second object, a distance between the first position information of each first object and the second position information of the second object in the same category as the first object, the method includes, before determining, for each first object, whether an angle between the first object and the second object in the same category with the smallest distance from the first object is smaller than a preset angle:
If the distance between the first object and the second object is larger than the set distance threshold value, deleting the identified distance, and executing the subsequent operation of judging whether the angle between the first object and the second object of the same category with the smallest distance between the first object and the second object is smaller than the preset angle.
In one possible embodiment, the method further comprises:
for each first object, judging whether a second object corresponding to the first object is identical to a second object corresponding to other first objects, and if so, re-determining the second object for the first object and other first objects with the same second object as the first object.
In a possible implementation manner, the determining, for a set number of image frames before the current frame image, second location information of each second object in the image frames, the method further includes:
determining a third object which is blocked in the current frame according to the track of each object determined in the image frames of the set number before the current frame;
and determining third position information of the third object in the current frame image according to the track of each object in the video detection frame before the current frame.
In one possible implementation, the traffic anomaly identification model is trained by:
for each sample image in a sample training set, determining a vector corresponding to the sample image according to sample first position information of each sample first object in the sample image and sample second position information of a second object corresponding to the sample first object in a set number of sample images before the current sample image;
determining a corresponding output value based on the vector through a traffic anomaly identification model;
and adjusting parameters of the traffic abnormality recognition model according to the output value.
In one possible implementation manner, the adjusting the parameters of the traffic anomaly identification model according to the output value includes:
according toξ i ＝max(0,1-O(G 2 (H(G′ 1 (X)))), adjusting parameters of the traffic anomaly identification model, wherein W 1 Is a first parameter, W 2 N is the number of vectors commonly input to the traffic anomaly recognition model, O (G 2 (H(G′ 1 (X)) is an output value corresponding to the vector, and r is a preset radius of the dividing plane.
Since the principle of the electronic device for solving the problem is similar to that of the communication method, the implementation of the electronic device can refer to the implementation of the method, and the repetition is omitted.
The communication bus mentioned above for the electronic devices may be a peripheral component interconnect standard (Peripheral Component Interconnect, PCI) bus or an extended industry standard architecture (Extended Industry Standard Architecture, EISA) bus, etc. The communication bus may be classified as an address bus, a data bus, a control bus, or the like. For ease of illustration, the figures are shown with only one bold line, but not with only one bus or one type of bus.
The communication interface 602 is used for communication between the electronic device and other devices described above.
The Memory may include random access Memory (Random Access Memory, RAM) or may include Non-Volatile Memory (NVM), such as at least one disk Memory. Optionally, the memory may also be at least one memory device located remotely from the aforementioned processor.
The processor may be a general-purpose processor, including a central processing unit, a network processor (Network Processor, NP), etc.; but also digital instruction processors (Digital Signal Processing, DSP), application specific integrated circuits, field programmable gate arrays or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.
Example 9:
on the basis of the above embodiments, the embodiments of the present invention further provide a computer readable storage medium having stored therein a computer program executable by a processor, which when run on the processor, causes the processor to perform the steps of:
acquiring first position information of each first object in the current frame image through a multi-target detection model which is completed through training;
determining second position information of each second object in the image frame according to the first position information of each first object and the second position information of the second object, wherein the second object corresponds to each first object of the current frame image in the image frame;
and determining whether the detection point is abnormal in traffic at the current frame moment based on the first position information of each first object in the current frame image and the second position information of the corresponding second object in the set number of image frames of the first object before the current frame image through the trained traffic abnormality recognition model.
In one possible embodiment, the first object comprises at least one of: pedestrians, bicycles, private cars, motorcycles, buses, and trucks.
In one possible implementation manner, the determining, according to the first position information of each first object and the second position information of the second object, the second object corresponding to each first object in the current frame image in the image frame includes:
determining the distance and the angle between the first position information of each first object and the second position information of the second object in the same category of the first object according to the first position information of each first object and the second position information of the second object;
for each first object, judging whether the angle between the first object and a second object of the same class with the smallest distance from the first object is smaller than a preset angle threshold value or not; if yes, determining the second object as a second object corresponding to the first object; if not, deleting the distance with the minimum distance, and re-determining whether the angle between the first object and the second object with the same category with the minimum distance with the first object is smaller than the preset angle threshold value.
In one possible implementation manner, after determining, according to the first position information of each first object and the second position information of the second object, a distance between the first position information of each first object and the second position information of the second object in the same category as the first object, the method includes, before determining, for each first object, whether an angle between the first object and the second object in the same category with the smallest distance from the first object is smaller than a preset angle:
If the distance between the first object and the second object is larger than the set distance threshold value, deleting the identified distance, and executing the subsequent operation of judging whether the angle between the first object and the second object of the same category with the smallest distance between the first object and the second object is smaller than the preset angle.
In one possible embodiment, the method further comprises:
for each first object, judging whether a second object corresponding to the first object is identical to a second object corresponding to other first objects, and if so, re-determining the second object for the first object and other first objects with the same second object as the first object.
In a possible implementation manner, the determining, for a set number of image frames before the current frame image, second location information of each second object in the image frames, the method further includes:
determining a third object which is blocked in the current frame according to the track of each object determined in the image frames of the set number before the current frame;
and determining third position information of the third object in the current frame image according to the track of each object in the video detection frame before the current frame.
In one possible implementation, the traffic anomaly identification model is trained by:
for each sample image in a sample training set, determining a vector corresponding to the sample image according to sample first position information of each sample first object in the sample image and sample second position information of a second object corresponding to the sample first object in a set number of sample images before the current sample image;
determining a corresponding output value based on the vector through a traffic anomaly identification model;
and adjusting parameters of the traffic abnormality recognition model according to the output value.
In one possible implementation manner, the adjusting the parameters of the traffic anomaly identification model according to the output value includes:
according toξ i ＝max(0,1-O(G 2 (H(G′ 1 (X)))), adjusting parameters of the traffic anomaly identification model, wherein W 1 Is a first parameter, W 2 N is the number of vectors commonly input to the traffic anomaly recognition model, O (G 2 (H(G′ 1 (X)) is an output value corresponding to the vector, and r is a preset radius of the dividing plane.
Since the principle of solving the problem with the computer readable medium provided above is similar to that of the communication method, the steps implemented after the processor executes the computer program in the computer readable medium can be referred to the other embodiments, and the repetition is omitted.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It will be apparent to those skilled in the art that various modifications and variations can be made to the present invention without departing from the spirit or scope of the invention. Thus, it is intended that the present invention also include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.
Claims (8)
1. A traffic anomaly determination method, the method comprising:
acquiring first position information of each first object in the current frame image through a multi-target detection model which is completed through training;
determining second position information of each second object in the image frame according to the first position information of each first object and the second position information of the second object, wherein the second object corresponds to each first object of the current frame image in the image frame;
Determining whether a traffic abnormality occurs at a detection point at the current frame moment based on first position information of each first object in the current frame image and second position information of a corresponding second object in a set number of image frames of the first object before the current frame image through a trained traffic abnormality identification model;
the determining, according to the first position information of each first object and the second position information of the second object, the second object corresponding to each first object of the current frame image in the image frame includes:
determining the distance and the angle between the first position information of each first object and the second position information of the second object in the same category of the first object according to the first position information of each first object and the second position information of the second object;
for each first object, judging whether the angle between the first object and the second object with the same category and the minimum distance between the first object is smaller than a preset angle threshold value or not; if yes, determining the second object as a second object corresponding to the first object; if not, deleting the second object with the smallest distance, and re-determining whether the angle between the first object and the second object with the same category with the smallest distance from the first object is smaller than a preset angle threshold value;
The traffic anomaly recognition model is trained by:
for each sample image in the sample training set, determining a vector corresponding to the sample image according to sample first position information of each sample first object in the sample image and sample second position information of a corresponding second object in a set number of sample images of the sample first object before the current sample image;
determining a corresponding output value based on the vector through a traffic anomaly identification model;
according to the output value, adjusting parameters of the traffic abnormality recognition model;
the traffic anomaly identification model comprises a first hidden layer and a second hidden layer; the first hidden layer is used for learning the linear characteristics of the vector in the training process of the traffic anomaly identification model; the second hidden layer is used for learning the nonlinear characteristics of the vector.
2. The method of claim 1, wherein the first object comprises at least one of: pedestrians, bicycles, private cars, motorcycles, buses, and trucks.
3. The method according to claim 1, wherein after determining the distance between the first position information of each first object and the second position information of the second object in the same category of the first object according to the first position information of each first object and the second position information of the second object, the method includes, before determining, for each first object, whether an angle between the first object and the second object in the same category, where the distance between the first object and the first object is the smallest, is smaller than a preset angle:
If the distance between the first object and the second object is larger than the set distance threshold value, deleting the identified distance, and executing the subsequent operation of judging whether the angle between the first object and the second object of the same class with the smallest distance between the first object and the second object is smaller than the preset angle.
4. A method according to claim 3, characterized in that the method further comprises:
for each first object, judging whether a second object corresponding to the first object is identical to a second object corresponding to other first objects, and if so, re-determining the second object for the first object and other first objects with the same second object as the first object.
5. The method of claim 1, wherein the second location information for each second object in the image frame is determined for a set number of image frames preceding the current frame image, the method further comprising:
determining a third object which is blocked in the current frame according to the track of each object determined in the image frames of the set number before the current frame;
and determining third position information of the third object in the current frame image according to the track of each object in the video detection frame before the current frame.
6. The method of claim 1, wherein said adjusting parameters of said traffic anomaly identification model based on said output values comprises:
according toξ i ＝max(0,1-O(G 2 (H(G 1 ′ (X)))), adjusting parameters of the traffic anomaly identification model, wherein W 1 Is a first parameter, W 2 N is the number of vectors commonly input to the traffic anomaly recognition model, O (G 2 (H(G 1 ′ (X)) is an output value corresponding to the vector, and γ is a parameter value set in advance.
7. An electronic device comprising at least a processor and a memory, the processor being adapted to perform the steps of the traffic anomaly determination method of any one of claims 1-6 when executing a computer program stored in the memory.
8. A computer-readable storage medium, characterized in that it stores a computer program which, when executed by a processor, performs the steps of the traffic anomaly determination method of any one of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011158286.8A CN112232257B (en) | 2020-10-26 | 2020-10-26 | Traffic abnormality determination method, device, equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011158286.8A CN112232257B (en) | 2020-10-26 | 2020-10-26 | Traffic abnormality determination method, device, equipment and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112232257A CN112232257A (en) | 2021-01-15 |
| CN112232257B true CN112232257B (en) | 2023-08-11 |
Family
ID=74109514
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011158286.8A Active CN112232257B (en) | 2020-10-26 | 2020-10-26 | Traffic abnormality determination method, device, equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112232257B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111047908B (en) * | 2018-10-12 | 2021-11-02 | 富士通株式会社 | Detection device and method for cross-line vehicle and video monitoring equipment |
| CN112528970A (en) * | 2021-02-07 | 2021-03-19 | 禾多科技（北京）有限公司 | Guideboard detection method, device, equipment and computer readable medium |
Citations (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103985257A (en) * | 2014-05-14 | 2014-08-13 | 南通大学 | Intelligent traffic video analysis method |
| CN104008390A (en) * | 2013-02-25 | 2014-08-27 | 三星泰科威株式会社 | Method and apparatus for detecting abnormal movement |
| CN105335986A (en) * | 2015-09-10 | 2016-02-17 | 西安电子科技大学 | Characteristic matching and MeanShift algorithm-based target tracking method |
| CN105488815A (en) * | 2015-11-26 | 2016-04-13 | 北京航空航天大学 | Real-time object tracking method capable of supporting target size change |
| CN106407946A (en) * | 2016-09-29 | 2017-02-15 | 北京市商汤科技开发有限公司 | Cross-line counting method, deep neural network training method, devices and electronic apparatus |
| CN106652465A (en) * | 2016-11-15 | 2017-05-10 | 成都通甲优博科技有限责任公司 | Method and system for identifying abnormal driving behavior on road |
| CN107766808A (en) * | 2017-09-30 | 2018-03-06 | 北京泓达九通科技发展有限公司 | The method and system that Vehicle Object motion track clusters in road network space |
| CN107820003A (en) * | 2017-09-28 | 2018-03-20 | 联想(北京)有限公司 | A kind of electronic equipment and control method |
| CN109087510A (en) * | 2018-09-29 | 2018-12-25 | 讯飞智元信息科技有限公司 | traffic monitoring method and device |
| CN109508576A (en) * | 2017-09-14 | 2019-03-22 | 杭州海康威视数字技术股份有限公司 | A kind of abnormal driving behavioral value method, apparatus and electronic equipment |
| CN109886995A (en) * | 2019-01-15 | 2019-06-14 | 深圳职业技术学院 | Multi-object tracking method under a kind of complex environment |
| CN109977833A (en) * | 2019-03-19 | 2019-07-05 | 网易（杭州）网络有限公司 | Object tracking method, object tracking device, storage medium and electronic equipment |
| CN110414313A (en) * | 2019-06-06 | 2019-11-05 | 平安科技（深圳）有限公司 | Abnormal behaviour alarm method, device, server and storage medium |
| CN110569785A (en) * | 2019-09-05 | 2019-12-13 | 杭州立宸科技有限公司 | Face recognition method based on fusion tracking technology |
| CN110991283A (en) * | 2019-11-21 | 2020-04-10 | 北京格灵深瞳信息技术有限公司 | Re-recognition and training data acquisition method and device, electronic equipment and storage medium |
| CN111047908A (en) * | 2018-10-12 | 2020-04-21 | 富士通株式会社 | Detection device and method for cross-line vehicle and video monitoring equipment |
| CN111081016A (en) * | 2019-12-18 | 2020-04-28 | 北京航空航天大学 | Urban traffic abnormity identification method based on complex network theory |
| CN111192454A (en) * | 2020-01-07 | 2020-05-22 | 中山大学 | Traffic abnormity identification method and system based on travel time evolution and storage medium |
| CN111260689A (en) * | 2020-01-16 | 2020-06-09 | 东华大学 | Effective confidence enhancement correlation filtering visual tracking algorithm |
| CN111753724A (en) * | 2020-06-24 | 2020-10-09 | 上海依图网络科技有限公司 | Abnormal behavior identification method and device |
| CN111814648A (en) * | 2020-06-30 | 2020-10-23 | 北京百度网讯科技有限公司 | Station port congestion situation determination method, device, equipment and storage medium |
-
2020
- 2020-10-26 CN CN202011158286.8A patent/CN112232257B/en active Active
Patent Citations (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104008390A (en) * | 2013-02-25 | 2014-08-27 | 三星泰科威株式会社 | Method and apparatus for detecting abnormal movement |
| CN103985257A (en) * | 2014-05-14 | 2014-08-13 | 南通大学 | Intelligent traffic video analysis method |
| CN105335986A (en) * | 2015-09-10 | 2016-02-17 | 西安电子科技大学 | Characteristic matching and MeanShift algorithm-based target tracking method |
| CN105488815A (en) * | 2015-11-26 | 2016-04-13 | 北京航空航天大学 | Real-time object tracking method capable of supporting target size change |
| CN106407946A (en) * | 2016-09-29 | 2017-02-15 | 北京市商汤科技开发有限公司 | Cross-line counting method, deep neural network training method, devices and electronic apparatus |
| CN106652465A (en) * | 2016-11-15 | 2017-05-10 | 成都通甲优博科技有限责任公司 | Method and system for identifying abnormal driving behavior on road |
| CN109508576A (en) * | 2017-09-14 | 2019-03-22 | 杭州海康威视数字技术股份有限公司 | A kind of abnormal driving behavioral value method, apparatus and electronic equipment |
| CN107820003A (en) * | 2017-09-28 | 2018-03-20 | 联想(北京)有限公司 | A kind of electronic equipment and control method |
| CN107766808A (en) * | 2017-09-30 | 2018-03-06 | 北京泓达九通科技发展有限公司 | The method and system that Vehicle Object motion track clusters in road network space |
| CN109087510A (en) * | 2018-09-29 | 2018-12-25 | 讯飞智元信息科技有限公司 | traffic monitoring method and device |
| CN111047908A (en) * | 2018-10-12 | 2020-04-21 | 富士通株式会社 | Detection device and method for cross-line vehicle and video monitoring equipment |
| CN109886995A (en) * | 2019-01-15 | 2019-06-14 | 深圳职业技术学院 | Multi-object tracking method under a kind of complex environment |
| CN109977833A (en) * | 2019-03-19 | 2019-07-05 | 网易（杭州）网络有限公司 | Object tracking method, object tracking device, storage medium and electronic equipment |
| CN110414313A (en) * | 2019-06-06 | 2019-11-05 | 平安科技（深圳）有限公司 | Abnormal behaviour alarm method, device, server and storage medium |
| CN110569785A (en) * | 2019-09-05 | 2019-12-13 | 杭州立宸科技有限公司 | Face recognition method based on fusion tracking technology |
| CN110991283A (en) * | 2019-11-21 | 2020-04-10 | 北京格灵深瞳信息技术有限公司 | Re-recognition and training data acquisition method and device, electronic equipment and storage medium |
| CN111081016A (en) * | 2019-12-18 | 2020-04-28 | 北京航空航天大学 | Urban traffic abnormity identification method based on complex network theory |
| CN111192454A (en) * | 2020-01-07 | 2020-05-22 | 中山大学 | Traffic abnormity identification method and system based on travel time evolution and storage medium |
| CN111260689A (en) * | 2020-01-16 | 2020-06-09 | 东华大学 | Effective confidence enhancement correlation filtering visual tracking algorithm |
| CN111753724A (en) * | 2020-06-24 | 2020-10-09 | 上海依图网络科技有限公司 | Abnormal behavior identification method and device |
| CN111814648A (en) * | 2020-06-30 | 2020-10-23 | 北京百度网讯科技有限公司 | Station port congestion situation determination method, device, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112232257A (en) | 2021-01-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112069944B (en) | Road congestion level determining method | |
| CN111310583B (en) | Vehicle abnormal behavior identification method based on improved long-term and short-term memory network | |
| CN112700470B (en) | Target detection and track extraction method based on traffic video stream | |
| CN104616502B (en) | Car license recognition and alignment system based on combination type bus or train route video network | |
| CN106652465A (en) | Method and system for identifying abnormal driving behavior on road | |
| CN112232257B (en) | Traffic abnormality determination method, device, equipment and medium | |
| CN108638999A (en) | A kind of collision early warning system and method for looking around input based on 360 degree | |
| CN110992693A (en) | Deep learning-based traffic congestion degree multi-dimensional analysis method | |
| CN102855500A (en) | Haar and HoG characteristic based preceding car detection method | |
| CN106650730A (en) | Turn signal lamp detection method and system in car lane change process | |
| CN113936465B (en) | Traffic event detection method and device | |
| CN112242058B (en) | Target abnormity detection method and device based on traffic monitoring video and storage medium | |
| Tourani et al. | Motion-based vehicle speed measurement for intelligent transportation systems | |
| CN113450573A (en) | Traffic monitoring method and traffic monitoring system based on unmanned aerial vehicle image recognition | |
| CN108445866B (en) | LDW false negative report test method and test system based on convolutional neural network | |
| CN111967396A (en) | Processing method, device and equipment for obstacle detection and storage medium | |
| CN108520528A (en) | Based on the mobile vehicle tracking for improving differential threshold and displacement field match model | |
| CN113361299B (en) | Abnormal parking detection method and device, storage medium and electronic equipment | |
| CN103310207B (en) | A kind of vehicle using motor detection method based on many Gauss models | |
| CN103258433B (en) | Intelligent clear display method for number plates in traffic video surveillance | |
| CN114141022B (en) | Emergency lane occupation behavior detection method and device, electronic equipment and storage medium | |
| CN112200835B (en) | Traffic accident detection method and device, electronic equipment and storage medium | |
| CN113158922A (en) | Traffic flow statistical method, device and equipment based on YOLO neural network | |
| CN112464749A (en) | Traffic scene abnormal target detection method based on rules and learning | |
| Castillo et al. | Vsion: Vehicle occlusion handling for traffic monitoring |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |