A full-time operational situation awareness system for traffic

Abstract

The invention discloses a full-time domain operational situation awareness system for traffic, comprising an optical fiber sensing acoustic wave demodulation module, a front-end computing power module, a display platform and a traffic camera system. The traffic camera system is used for real-time monitoring of road traffic conditions, collecting traffic data and assisting traffic law enforcement. The optical fiber sensing acoustic wave demodulation module is used for connecting optical fibers for acoustic wave data collection. The front-end computing power module is used for connecting the optical fiber sensing acoustic wave demodulation module to receive the collected acoustic wave data and to run a data processing algorithm in real time to extract vehicle information. The display platform is used for receiving vehicle information obtained by the front-end computing power module, analyzing existing abnormal events and linking with the traffic camera system. Through the implementation of the invention, a series of algorithms and technical means for effectively identifying and separating vehicle trajectories in complex scenarios are realized, and vehicle trajectories in videos or images are accurately identified and tracked, thereby realizing monitoring and management of vehicle motion states, and having certain use value and promotion value.

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation and data processing technology, and in particular to a full-time domain operation situation awareness system for transportation. Background Art

[0002] An ideal smart highway system should have high-precision holographic perception of all highway elements, highly reliable information exchange, and highly autonomous intelligent traffic control capabilities. In other words, it should be able to sense the movement status of each vehicle in all weather and on all roads, and be able to accurately control vehicles in various application scenarios. To achieve this goal, it is necessary to find a basic technical means that can achieve full-time domain perception.

[0003] At present, the perception means of highways and urban elevated roads are mainly video surveillance, which is easily affected by bad weather and lighting conditions. The perception ability is greatly reduced in rainy, foggy and backlit conditions. In addition, due to the different construction stages, different technical means and complex application scenarios of highway perception equipment, the data forms are different, the application value is different, and the integration difficulty is huge, which makes it difficult to support the digital upgrade of the perception layer.

[0004] In summary, a full-time operational situation awareness system for traffic is needed to address the shortcomings of existing technologies. Summary of the invention

[0005] In view of the shortcomings of the existing technology, the present invention provides a full-time domain operational situation awareness system for traffic, which aims to solve the problem that the current perception of highways is greatly affected by weather, the data of perception devices are in different forms, the application value is different, and the integration is difficult, which makes it difficult to support the digital upgrade of the perception layer.

[0006] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: a full-time domain operational situation awareness system for traffic, including a fiber optic sensing acoustic wave demodulation module, a front-end computing power module, a display platform and a traffic camera system, wherein the traffic camera system is used to monitor road traffic conditions in real time, collect traffic data, and assist in traffic law enforcement, wherein the fiber optic sensing acoustic wave demodulation module is used to connect the optical fiber for acoustic wave data collection, wherein the front-end computing power module is used to connect the fiber optic sensing acoustic wave demodulation module to receive the collected acoustic wave data and to run the data processing algorithm in real time to extract vehicle information, wherein the display platform is used to receive the vehicle information obtained by the front-end computing power module, analyze the existing abnormal events and link with the traffic camera system. Through the integrated use of the fiber optic sensing acoustic wave demodulation module, the front-end computing power module and the display platform, a series of algorithms and technical means for effectively identifying and separating vehicle trajectories in complex scenarios are realized, and the vehicle trajectories in videos or images are accurately identified and tracked, thereby realizing the monitoring and management of the vehicle motion state.

[0007] Furthermore, the front-end computing module extracts vehicle information through a data processing algorithm, and the data processing algorithm includes the following steps:

[0008] Step 1: Use image annotation tools to mark each vehicle spatiotemporal trajectory in the original data to generate vehicle label data;

[0009] Step 2: Preprocess the vehicle label data and the original image to obtain the enhanced data block and input it into the deep learning network model as a training set;

[0010] Step 3: Use the deep learning network model to detect the target, identify the overall situation of the vehicle trajectory and extract feature data;

[0011] Step 4: Input the feature data extracted in step 3 into the clustering model to distinguish different vehicle trajectories;

[0012] Step 5: Input the different vehicle trajectories distinguished in step 4 into the vehicle separation algorithm to separate the trajectories of vehicles with intersecting directions;

[0013] Step 6: Generate a temporary vehicle point set P_m according to the separated trajectories of the crossing vehicles;

[0014] Step 7: Compare each point set in the temporary vehicle point set P_m (the number of vehicle point sets is M) with each point set in the existing vehicle point set P_n (the number of vehicle point sets is N) in a circular manner;

[0015] Step 8: During the cyclic comparison, when point set i in the temporary vehicle point set P_m belongs to point set j in the existing vehicle point set P_n, point set i will be added to point set j; when point set i in the temporary vehicle point set P_m does not belong to any point set in the vehicle point set P_n, point set i will be added to the existing vehicle point set P_n, and the number of vehicle point sets will be increased to N+1;

[0016] Among them, the vehicle point set set P_m represents that the number of vehicle point sets is M, and the vehicle point set set P_n represents that the number of vehicle point sets is N.

[0017] Furthermore, in step 4, the feature data is input into a clustering model to distinguish different vehicle trajectories and form different clusters.

[0018] Furthermore, the clustering model distinguishes different vehicle trajectories and forms different clusters, including the following steps:

[0019] Step A1: First input the binary image to be classified;

[0020] Step A2: Then get the x-coordinates and y-coordinates of all data points;

[0021] Step A3: Then perform DBScan clustering and obtain the category label list corresponding to all data points;

[0022] Step A4: Loop to obtain the category label, and assign the value of all data points corresponding to the category label in the binary image to i+1;

[0023] Step A5: After completing the above steps, the clustered data is returned;

[0024] Step A6: According to the returned clustering data, loop to obtain the corresponding point set of the vehicles therein;

[0025] Step A7: judging whether there is an intersection by using a vehicle separation algorithm according to the corresponding point set obtained in step A6; if there is an intersection, processing the current point set; if there is no intersection, adding the current point set to the vehicle list;

[0026] Step A8: Finally, complete all vehicle processing procedures.

[0027] Furthermore, in step A7, the vehicle list includes all identified vehicles and their corresponding point sets.

[0028] Furthermore, in step 5, different vehicle trajectories are input into a vehicle separation algorithm to separate trajectories of vehicles with intersecting directions, and the trajectories of vehicles with intersecting directions include trajectories of vehicles with different traveling directions.

[0029] Furthermore, the method for separating the trajectories of vehicles with intersecting directions using the vehicle separation algorithm comprises the following steps:

[0030] Step B1: Regenerate the image based on the vehicle point set obtained by the DBScan algorithm as the algorithm input;

[0031] Step B2: Initialize the vehicle point set list;

[0032] Step B3: Calculate the difference between adjacent data points in the x direction for the input image;

[0033] Step B4: extracting data with non-zero difference values ​​of adjacent data points starting from row 0, whose symbols represent the boundary of a vehicle point set;

[0034] Step B5: Get the left and right boundary x-coordinate lists left0, right0;

[0035] Step B6: Determine whether the lengths of the left and right boundary x-coordinate lists are equal, if not, execute step B7, if equal, execute step B8;

[0036] Step B7: append 0 to the list left0 or nx to the list right0 to make the length of left0 and right0 consistent;

[0037] Step B8: Find the midpoint coordinate list mid0;

[0038] Step B9: Obtain the vehicle point set and index in the current row according to mid0 and add them to the vehicle point set set and index list, the number of vehicles is N;

[0039] Step B10: Then, obtain data through a row loop, perform the above processing on each i-th row (i>0) of data to obtain a midpoint coordinate list mid_i, and then compare it with i-1;

[0040] Step B11: Determine whether the number of elements in mid_i is equal to that in mid_i-1. If they are not equal, execute step B16. If they are equal, execute step B12.

[0041] Step B12: Through a double loop, compare the idx1th element in mid_i with the idxth element in mid_i-1;

[0042] Step B13: When the idx1th element in mid_i can find any element in mid_i-1 that satisfies abs(mid_i[idx1]-mid_i-1[idx])<=3, if not, execute step B16; if yes, execute step B14;

[0043] Step B14: Calculate the slope k of the vehicle point set corresponding to mid_i[idx1] and mid_i-1[idx] and compare them. If the slopes are the same, add the point set in mid_i to the vehicle point set corresponding to mid_i-1;

[0044] Step B15: After the double loop is completed, update the midpoint coordinates and the left and right boundary x coordinates of the current row mid_i-1=mid_i, left_i-1=left_i, right_i-1=right_i;

[0045] Step B16: Determine whether the left boundary left_i[idx1] corresponding to the idx1th element in mid_i can find any left_i-1[idx] that satisfies the following conditions in mid_i-1,

[0046] abs(left_i-1[idx1]-left_i[idx])<3;

[0047] Determine whether the right boundary right_i[idx1] corresponding to the idx1th element in mid_i can find any right_i-1[idx1] that satisfies the following conditions in mid_i-1,

[0048] abs(right_i-1[idx1]-right_i[idx])<3;

[0049] If both right_i-1[idx1] and left_i-1[idx] exist, mid_i[idx1] is considered to be the intersection trajectory, and the process returns directly to step B12; if step B16 is not established, a new vehicle point set is added to the vehicle point set list, the number of vehicles becomes N+1, and then the process jumps to step B12.

[0050] Furthermore, in step 6, each point set in the temporary vehicle point set set P_m is cyclically compared with each point set in the existing vehicle point set set P_n, and the content of the cyclic comparison includes the slope and the position of the start-end point.

[0051] Furthermore, in step 8, the temporary vehicle point set i is added to the existing vehicle point set j according to the result of the circular comparison, and the number of vehicle point sets is increased to N+1 according to the judgment situation, where l is the number of vehicles to be added.

[0052] The substantial effect of the present invention is as follows: a vehicle separation model is used to make a lateral difference in the original data to obtain the left and right contours of each signal, and then the left and right contours are matched by row traversal. When an intersection occurs, multiple pairs of left and right contours will appear, and the number of left and right contours at the intersection will definitely decrease. Through this conditional judgment, the intersection point set can be separated, thereby distinguishing the traffic flows in different directions. The extracted vehicle trajectory can be analyzed to obtain information such as the driving direction, vehicle speed, and the position of the meter mark on the optical cable. Through the correction method, the meter mark on the optical cable can be matched one by one with the K mark in traffic management and maintenance. The data information is transmitted to the display platform through the network and is stored in the database. At the same time, the vehicle information is synchronized to the platform system for real-time warning and updating. When an abnormal event occurs, the K mark obtained after correction can be used to dispatch the existing camera system through the system to find the camera closest to the event location, and the accident can be reviewed based on the image video. In addition, the change in vehicle speed can also be used to determine whether there are abnormal events such as abnormal parking and traffic accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a system principle block diagram of the present invention.

[0054] Figure 2 It is a schematic diagram of the process of the present invention.

[0055] Figure 3Schematic diagram of vehicle trajectory extracted by Yolov8 of the present invention.

[0056] Figure 4 A schematic diagram of generating vehicle tag data according to the present invention. DETAILED DESCRIPTION

[0057] In order to facilitate the understanding of the present invention, the present invention is described in more detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that when an element is described as "fixed to" another element, it can be directly on the other element, or there can be one or more centered elements therebetween. When an element is described as "connected to" another element, it can be directly connected to the other element, or there can be one or more centered elements therebetween. The terms "vertical", "horizontal", "left", "right" and similar expressions used in this specification are for illustrative purposes only.

[0058] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used in this specification includes any and all combinations of one or more of the related listed items.

[0059] Reference Figure 1 A full-time domain operational situation awareness system for traffic, including a fiber optic sensing acoustic wave demodulation module, a front-end computing power module, a display platform and a traffic camera system. The traffic camera system is used to monitor road traffic conditions in real time, collect traffic data, and assist in traffic law enforcement. The fiber optic sensing acoustic wave demodulation module is used to connect the optical fiber for acoustic wave data collection. The front-end computing power module is used to connect the fiber optic sensing acoustic wave demodulation module to receive the collected acoustic wave data and run the data processing algorithm in real time to extract vehicle information. The display platform is used to receive the vehicle information obtained by the front-end computing power module, analyze the existing abnormal events and link with the traffic camera system. The abnormal events include abnormal parking on the curb, abnormal parking on the lane, collision events and traffic congestion, etc. The traffic camera system is an existing traffic camera system.

[0060] Reference Figure 2 , the front-end computing module extracts vehicle information through a data processing algorithm, which includes the following steps:

[0061] Step 1: Use image annotation tools to mark each vehicle spatiotemporal trajectory in the original data to generate vehicle label data;

[0062] Step 2: Preprocess the vehicle label data and the original image to obtain the enhanced data block and input it into the Yolov8 deep learning network model as a training set;

[0063] Step 3: Use the Yolov8 deep learning network model to detect targets, identify the overall situation of vehicle trajectories and extract feature data (refer to Figure 3 );

[0064] Step 4: Input the feature data extracted in step 3 into the clustering model to distinguish different vehicle trajectories;

[0065] Step 5: Input the different vehicle trajectories distinguished in step 4 into the vehicle separation algorithm to separate the trajectories of vehicles with intersecting directions;

[0066] Step 6: Generate a temporary vehicle point set P_m according to the separated trajectories of the crossing vehicles;

[0067] Step 7: Compare each point set in the temporary vehicle point set P_m (the number of vehicle point sets is M) with each point set in the existing vehicle point set P_n (the number of vehicle point sets is N) in a circular manner;

[0068] Step 8: During the cyclic comparison, when point set i in the temporary vehicle point set P_m belongs to point set j in the existing vehicle point set P_n, point set i will be added to point set j; when point set i in the temporary vehicle point set P_m does not belong to any point set in the vehicle point set P_n, point set i will be added to the existing vehicle point set P_n, and the number of vehicle point sets will be increased to N+1;

[0069] Among them, the vehicle point set set P_m represents that the number of vehicle point sets is M, and the vehicle point set set P_n represents that the number of vehicle point sets is N.

[0070] In step 4, the feature data is input into the clustering model to distinguish different vehicle trajectories and form different clusters.

[0071] The clustering model distinguishes different vehicle trajectories and forms different clusters, including the following steps:

[0072] Step A1: First input the binary image to be classified;

[0073] Step A2: Then get the x-coordinates and y-coordinates of all data points;

[0074] Step A3: Then perform DBScan clustering and obtain the category label list corresponding to all data points;

[0075] Step A4: Loop to obtain the category label, and assign the value of all data points corresponding to the category label in the binary image to i+1;

[0076] Step A5: After completing the above steps, the clustered data is returned;

[0077] Step A6: According to the returned clustering data, loop to obtain the corresponding point set of the vehicles therein;

[0078] Step A7: judging whether there is an intersection by using a vehicle separation algorithm according to the corresponding point set obtained in step A6; if there is an intersection, processing the current point set; if there is no intersection, adding the current point set to the vehicle list;

[0079] Step A8: Finally, complete all vehicle processing procedures.

[0080] In step A7, the vehicle list includes all identified vehicles and their corresponding point sets.

[0081] DBScan is a clustering algorithm based on sample density. Samples of the same category are closely connected. By classifying closely connected samples into one category, a cluster category can be obtained. DBScan describes the compactness of a sample set based on a set of neighborhoods. The parameter (e, MinPts) is used to describe the compactness of sample distribution in the neighborhood. Among them, e describes the neighborhood distance threshold of a certain sample, and MinPts describes the threshold of the number of samples in the neighborhood of a certain sample with a distance of e. It can effectively process data point sets of any shape, does not require the pre-specified number of clusters, and has good robustness to noise points.

[0082] The DBScan algorithm consists of the following steps:

[0083] Step C1: Randomly select a data point p that violates access;

[0084] Step C2: Taking p as the center and e as the radius, find all data points within the radius;

[0085] Step C3: If the number of data points within the radius is less than MinPts, p is marked as a noise point;

[0086] Step C4: Otherwise, create a new cluster with point p as the core and add all the points within the radius to the cluster;

[0087] Step C5: Take all points within the radius as seed points and repeat the above process until the cluster is completely discovered.

[0088] Step C6: Repeat the above process until no new clusters appear;

[0089] Step C7: The samples that are not finally classified into any cluster are called noise points and will be treated as separate categories or discarded.

[0090] In step 5, different vehicle trajectories are input into a vehicle separation algorithm to separate the trajectories of vehicles with intersecting directions, and the trajectories of vehicles with intersecting directions include the trajectories of vehicles with different traveling directions.

[0091] The method of separating the trajectories of vehicles with intersecting directions by the vehicle separation algorithm comprises the following steps:

[0092] Step B1: Regenerate the image based on the vehicle point set obtained by the DBScan algorithm as the algorithm input;

[0093] Step B2: Initialize the vehicle point set list;

[0094] Step B3: Calculate the difference between adjacent data points in the x direction for the input image;

[0095] Step B4: extracting data with non-zero difference values ​​of adjacent data points starting from row 0, whose symbols represent the boundary of a vehicle point set;

[0096] Step B5: Get the left and right boundary x-coordinate lists left0, right0;

[0097] Step B6: Determine whether the lengths of the left and right boundary x-coordinate lists are equal, if not, execute step B7, if equal, execute step B8;

[0098] Step B7: append 0 to the list left0 or nx to the list right0 to make the length of left0 and right0 consistent;

[0099] Step B8: Find the midpoint coordinate list mid0;

[0100] Step B9: Obtain the vehicle point set and index in the current row according to mid0 and add them to the vehicle point set set and index list, the number of vehicles is N;

[0101] Step B10: Then, obtain data through a row loop, perform the above processing on each i-th row (i>0) of data to obtain a midpoint coordinate list mid_i, and then compare it with i-1;

[0102] Step B11: Determine whether the number of elements in mid_i is equal to that in mid_i-1. If they are not equal, execute step B16. If they are equal, execute step B12.

[0103] Step B12: Through a double loop, compare the idx1th element in mid_i with the idxth element in mid_i-1;

[0104] Step B13: When the idx1th element in mid_i can find any element in mid_i-1 that satisfies abs(mid_i[idx1]-mid_i-1[idx])<=3, if not, execute step B16; if yes, execute step B14;

[0105] Step B14: Calculate the slope k of the vehicle point set corresponding to mid_i[idx1] and mid_i-1[idx] and compare them. If the slopes are the same, add the point set in mid_i to the vehicle point set corresponding to mid_i-1;

[0106] Step B15: After the double loop is completed, update the midpoint coordinates and the left and right boundary x coordinates of the current row mid_i-1=mid_i, left_i-1=left_i, right_i-1=right_i;

[0107] Step B16: Determine whether the left boundary left_i[idx1] corresponding to the idx1th element in mid_i can find any left_i-1[idx] that satisfies the following conditions in mid_i-1,

[0108] abs(left_i-1[idx1]-left_i[idx])<3;

[0109] Determine whether the right boundary right_i[idx1] corresponding to the idx1th element in mid_i can find any right_i-1[idx1] that satisfies the following conditions in mid_i-1,

[0110] abs(right_i-1[idx1]-right_i[idx])<3;

[0111] If both right_i-1[idx1] and left_i-1[idx] exist, mid_i[idx1] is considered to be the intersection trajectory, and the process returns directly to step B12; if step B16 is not established, a new vehicle point set is added to the vehicle point set list, the number of vehicles becomes N+1, and then the process jumps to step B12.

[0112] In step 6, each point set in the temporary vehicle point set set P_m is cyclically compared with each point set in the existing vehicle point set set P_n, and the content of the cyclic comparison includes the slope and the position of the start-end point;

[0113] The data stored in the vehicle point set is a two-dimensional array, which represents the coordinate position of the vehicle in time and space. When each newly detected vehicle is added to the vehicle list, its current time and space position will be predicted for the next time it will appear, and added to a prediction list. That is, the length of the prediction list is the same as the length of the vehicle list.

[0114] When performing a cyclic comparison between a temporary vehicle and the vehicle list, it will first determine whether the time and space position of the temporary vehicle can be matched in the prediction list. If not, it will be considered as a new vehicle and added to the vehicle list, and the prediction list will be updated.

[0115] If it can be matched in the prediction list, all corresponding vehicles are taken out from the vehicle list according to the index of the prediction list and compared with the current vehicle in a loop. The content of the comparison is the slope and the information of the starting and ending points. Because the two-dimensional data of the vehicle point set can be approximately represented as a line segment in the plane in geometry, the slope of the line and the starting-end point information can be obtained from the line segment. Therefore, when the slopes of the line segments represented by the two vehicles are equal and the starting point of one line segment is very close to the ending point of the other line segment, they are judged as one vehicle, otherwise they are different vehicles.

[0116] The steps for comparing the slope and the start-end point positions are as follows:

[0117] Determine the direction and difference of the slopes of the two vehicles. First, calculate the slopes of the two vehicles. When the product of the slopes of the line segments represented by the two vehicles is greater than zero (i.e., the slope sign is +, + or -, -) and the difference between the two slopes is less than a certain value, proceed to step 2, otherwise end the comparison;

[0118] To determine whether the starting point and the end point of two vehicles are close, first determine the positional relationship between the two vehicles. For example, compare the newly detected vehicle A with the vehicle B in the vehicle list. First, determine whether the y-axis coordinate of the starting point of A (i.e., the position in time) is greater than or equal to the y-axis coordinate of the end point of B (condition 1). If so, compare the x-axis coordinate of the starting point of A (i.e., the position in space) with the x-axis coordinate of the end point of B. When the difference is in a specified interval, AB is considered to be a vehicle. Otherwise, continue to loop and compare the next vehicle in the vehicle list; if the above condition 1 is not met, compare the positional relationship between the starting point of B and the end point of A in the same way (condition 2); if both conditions 1 and 2 are not met, exit the comparison and proceed to the next cycle; when the cycle ends, if A has not matched any vehicle in the vehicle list, A is considered to be a new vehicle and added to the vehicle list, and the prediction list is updated. Otherwise, the point set corresponding to A is added to the vehicle matched with it.

[0119] In step 8, the temporary vehicle point set i is added to the existing vehicle point set j according to the result of the circular comparison, and the number of vehicle point sets is increased to N+1 according to the judgment situation, where l is the number of vehicles to be added.

[0120] Due to the different directions of vehicles, the space-time trajectories of vehicles on the image will cross (refer to Figure 3), the intersecting trajectories are classified into one category in the DBScan clustering algorithm. The vehicle separation algorithm calculates the left and right contours of each signal by making a lateral difference on the original data, and then matches the left and right contours by row traversal. When an intersection occurs, multiple pairs of left and right contours will appear, and the number of left and right contours at the intersection will definitely decrease. Through this conditional judgment, the intersection point set can be separated, and the traffic in different directions can be distinguished.

[0121] The separated point sets will be assigned M different temporary vehicle IDs, and the temporary vehicle point sets will be compared with the existing N vehicle point sets in a loop, that is, the i-th temporary vehicle point set and the j-th existing vehicle point set will determine whether they belong to the same vehicle based on the speed and the start-end point positions. If they match, the i-th temporary vehicle point set will be assigned to the existing j-th vehicle point set. If there is no match in the existing vehicle point set, the temporary vehicle point set will be added to the existing vehicle point set as a whole, and the number of existing vehicle point sets will become N=N+l.

[0122] It should be noted that the preferred embodiments of the present invention are given in the specification and drawings of the present invention. However, the present invention can be implemented in many different forms and is not limited to the embodiments described in this specification. These embodiments are not intended to be additional limitations on the content of the present invention. The purpose of providing these embodiments is to make the understanding of the disclosure of the present invention more thorough and comprehensive. In addition, the above-mentioned technical features continue to be combined with each other to form various embodiments not listed above, which are all considered to be within the scope of the present invention; further, for ordinary technicians in this field, they can be improved or transformed according to the above description, and all these improvements and transformations should belong to the scope of protection of the claims attached to the present invention.

Claims

1. A full-time operational situation awareness system for traffic, comprising an optical fiber sensing acoustic wave demodulation module, a front-end computing module, a display platform and a traffic camera system, wherein the traffic camera system is used to monitor road traffic conditions in real time, collect traffic data, and assist in traffic law enforcement, and is characterized in that: The optical fiber sensing acoustic wave demodulation module is used to connect to the optical fiber for acoustic wave data collection, the front-end computing module is used to connect to the optical fiber sensing acoustic wave demodulation module to receive the collected acoustic wave data and run the data processing algorithm in real time to extract vehicle information, and the display platform is used to receive the vehicle information obtained by the front-end computing module, analyze the existing abnormal events and link with the traffic camera system; The front-end computing module extracts vehicle information through a data processing algorithm, and the data processing algorithm includes the following steps: Step 1: Use image annotation tools to mark each vehicle spatiotemporal trajectory in the original data to generate vehicle label data; Step 2: Preprocess the vehicle label data and the original image to obtain the enhanced data block and input it into the deep learning network model as a training set; Step 3: Use the deep learning network model to detect the target, identify the overall situation of the vehicle trajectory and extract feature data; Step 4: Input the feature data extracted in step 3 into the clustering model to distinguish different vehicle trajectories; Step 5: Input the different vehicle trajectories distinguished in step 4 into the vehicle separation algorithm to separate the trajectories of vehicles with intersecting directions; Step 6: Generate a temporary vehicle point set P_m according to the separated trajectories of the crossing vehicles; Step 7: Compare each point set in the temporary vehicle point set P_m with each point set in the existing vehicle point set P_n in a circular manner; Step 8: During the cyclic comparison, when point set i in the temporary vehicle point set P_m belongs to point set j in the existing vehicle point set P_n, point set i will be added to point set j; when point set i in the temporary vehicle point set P_m does not belong to any point set in the vehicle point set P_n, point set i will be added to the existing vehicle point set P_n, and the number of vehicle point sets will be increased to N+1; Among them, the vehicle point set set P_m represents that the number of vehicle point sets is M, and the vehicle point set set P_n represents that the number of vehicle point sets is N.

2. A full-time operational situation awareness system for traffic according to claim 1, characterized in that: In step 4, the feature data is input into the clustering model to distinguish different vehicle trajectories and form different clusters.

3. A full-time operational situation awareness system for traffic according to claim 2, characterized in that: The clustering model distinguishes different vehicle trajectories and forms different clusters, comprising the following steps: Step A1: First input the binary image to be classified; Step A2: Then get the x-coordinates and y-coordinates of all data points; Step A3: Then perform DBScan clustering and obtain the category label list corresponding to all data points; Step A4: Loop to obtain the category label, and assign the value of all data points corresponding to the category label in the binary image to i+1; Step A5: After completing the above steps, the clustered data is returned; Step A6: According to the returned clustering data, loop to obtain the corresponding point set of the vehicles therein; Step A7: judging whether there is an intersection by using a vehicle separation algorithm according to the corresponding point set obtained in step A6; if there is an intersection, processing the current point set; if there is no intersection, adding the current point set to the vehicle list; Step A8: Finally, complete all vehicle processing procedures.

4. A full-time operational situation awareness system for traffic according to claim 3, characterized in that: In step A7, the vehicle list includes all identified vehicles and their corresponding point sets.

5. The full-time operational situation awareness system for traffic according to claim 1, characterized in that: In step 5, different vehicle trajectories are input into a vehicle separation algorithm to separate trajectories of vehicles with intersecting directions, and the trajectories of vehicles with intersecting directions include trajectories of vehicles with different traveling directions.

6. A full-time operational situation awareness system for traffic according to claim 5, characterized in that: The method for separating the trajectories of vehicles with intersecting directions by the vehicle separation algorithm comprises the following steps: Step B1: Regenerate the image based on the vehicle point set obtained by the DBScan algorithm as the algorithm input; Step B2: Initialize the vehicle point set list; Step B3: Calculate the difference between adjacent data points in the x direction for the input image; Step B4: extracting data with non-zero difference values ​​of adjacent data points starting from row 0, whose symbols represent the boundary of a vehicle point set; Step B5: Get the left and right boundary x-coordinate lists left0, right0; Step B6: Determine whether the lengths of the left and right boundary x-coordinate lists are equal, if not, execute step B7, if equal, execute step B8; Step B7: append 0 to the list left0 or nx to the list right0 to make the length of left0 and right0 consistent; Step B8: Find the midpoint coordinate list mid0; Step B9: Obtain the vehicle point set and index in the current row according to mid0 and add them to the vehicle point set set and index list, the number of vehicles is N; Step B10: Then, obtain data through a row loop, perform the above processing on each i-th row (i>0) of data to obtain a midpoint coordinate list mid_i, and then compare it with i-1; Step B11: Determine whether the number of elements in mid_i is equal to that in mid_i-1. If they are not equal, execute step B16. If they are equal, execute step B12. Step B12: Through a double loop, compare the idx1th element in mid_i with the idxth element in mid_i-1; Step B13: When the idx1th element in mid_i can find any element in mid_i-1 that satisfies abs(mid_i[idx1]-mid_i-1[idx])<=3, if not, execute step B16; if yes, execute step B14; Step B14: Calculate the slope k of the vehicle point set corresponding to mid_i[idx1] and mid_i-1[idx] and compare them. If the slopes are the same, add the point set in mid_i to the vehicle point set corresponding to mid_i-1; Step B15: After the double loop is completed, update the midpoint coordinates and the left and right boundary x coordinates of the current row mid_i-1=mid_i, left_i-1=left_i, right_i-1=right_i; Step B16: Determine whether the left boundary left_i[idx1] corresponding to the idx1th element in mid_i can find any left_i-1[idx] that satisfies the following conditions in mid_i-1, abs(left_i-1[idx1]-left_i[idx])<3; Determine whether the right boundary right_i[idx1] corresponding to the idx1th element in mid_i can find any right_i-1[idx1] that satisfies the following conditions in mid_i-1, abs(right_i-1[idx1]-right_i[idx])<3; If both right_i-1[idx1] and left_i-1[idx] exist, mid_i[idx1] is considered to be the intersection trajectory, and the process returns directly to step B12; if step B16 is not established, a new vehicle point set is added to the vehicle point set list, the number of vehicles becomes N+1, and then the process jumps to step B12.

7. The full-time operational situation awareness system for traffic according to claim 1, characterized in that: In step 7, each point set in the temporary vehicle point set set P_m is cyclically compared with each point set in the existing vehicle point set set P_n, and the content of the cyclic comparison includes the slope and the position of the start-end point.

8. The full-time operational situation awareness system for traffic according to claim 1, characterized in that: In step 8, the temporary vehicle point set i is added to the existing vehicle point set j according to the result of the circular comparison, and the number of vehicle point sets is increased to N+1 according to the judgment situation, where l is the number of vehicles to be added.

Images