UAV multi-source sensing crowd intelligence evaluation method for sudden traffic congestion

Abstract

The present invention proposes a multi-source sensing group intelligence assessment method for sudden traffic congestion using drones. The method includes the following steps: first, an improved YOLOv5 aerial vehicle detection and tracking algorithm is proposed, and data preprocessing is performed on drone aerial images by performing vehicle bounding box detection based on improved K-MEANS clustering; then, a feature extraction network based on improved EFFICIENTMET is proposed to improve detection accuracy. Secondly, the motion state of the vehicle is calculated based on the altitude, viewing angle, posture, and position information of the drone, and the congestion state of the traffic section is quickly assessed using motion information such as the position, spacing, and speed between vehicles through multi-source sensing fusion technology. Under congestion conditions, the distance to the congested section is calculated based on the drone's field of view mapping, and the number of congestions is counted using an improved DeepSort algorithm to intelligently assess the congestion state of the traffic section. This achieves comprehensive assessment and monitoring of sudden traffic congestion, provides a more flexible and intelligent solution for traffic management, and improves the efficiency of urban traffic operations.

Description

Technical Field

[0001] The present invention relates to the technical field of autonomous perception, multi-source sensor fusion, and swarm intelligence decision-making of UAVs in complex mission contexts, and specifically to a UAV multi-source sensor swarm intelligence assessment method for sudden traffic congestion. Background Art

[0002] Existing traffic congestion assessment technologies, due to the randomness of time, space, and timing, are unable to quickly and accurately determine the degree of congestion, calculate vehicle density, obtain the real-time location of traffic jams, provide real-time monitoring and assessment, and rapidly respond to sudden traffic events. Due to the randomness of time, space, and timing, most algorithms are unable to quickly and accurately determine the degree of congestion, calculate vehicle density, obtain the real-time location of traffic jams, provide real-time monitoring and assessment, and respond quickly to sudden traffic events. Furthermore, due to the limited data source and slow update speed, traditional methods often have difficulty in timely detecting and predicting traffic congestion, which affects the efficiency of emergency response and traffic management. Summary of the Invention

[0003] Drones offer rapid response, flexibility, and wide coverage for sudden traffic congestion. They can provide a real-time, high-altitude perspective, quickly reach the scene, and obtain real-time traffic data, comprehensively monitoring traffic conditions and eliminating the limited coverage and blind spots inherent in traditional methods. Secondly, equipped with advanced sensing technology, drones can rapidly collect large amounts of traffic data, enabling more accurate congestion analysis and helping to promptly identify and address traffic issues. Drones offer the advantages of rapid deployment and maneuverability, allowing them to flexibly adapt to diverse traffic scenarios and improve the efficiency of congestion monitoring. Overall, the use of drones in traffic congestion assessment not only enhances the comprehensiveness and real-time nature of data collection, but also provides a more flexible and intelligent solution for traffic management, enabling the prediction and early warning of traffic congestion and improving the efficiency and sustainability of urban transportation operations.

[0004] The present invention proposes a multi-source sensing swarm intelligence assessment method for UAVs targeting sudden traffic congestion. UAVs are equipped with a variety of sensors, such as high-definition cameras, lidars, infrared sensors, etc., to achieve multi-source data collection and provide more comprehensive and accurate traffic congestion information. An improved YOLOv5 algorithm is used to detect aerial vehicles, and the vehicle's motion state is calculated based on the UAV's altitude, viewing angle, posture, and position information. Multi-source sensing fusion technology is used to perform swarm intelligence assessment of the position, spacing, speed and other motion information between vehicles, and a rapid judgment is made on traffic congestion. In the congested state, the distance to the congested section is calculated based on the UAV's field of view mapping of the ground, and the number of congestions is counted using an improved DeepSort algorithm to intelligently assess the congestion state of the traffic section. Through the multi-source sensing swarm intelligence assessment method for UAVs targeting sudden traffic congestion, the multi-source sensing data acquired by the UAV are integrated and analyzed to achieve a comprehensive assessment and monitoring of traffic congestion.

[0005] The specific process of the UAV multi-source sensing crowd intelligence assessment method for sudden traffic congestion is as follows:

[0006] Step 1: Use the improved YOLOv5 algorithm to detect the aerial vehicle target;

[0007] Step 2: Calculate the vehicle's motion state based on the drone's altitude, viewing angle, attitude, and position information;

[0008] Step 3: Use multi-source sensor fusion technology to perform group intelligence assessment of vehicle motion information and quickly determine whether a traffic section is congested.

[0009] Step 4: Use the improved DeepSort algorithm to count congested vehicles;

[0010] Step 5: Calculate the distance to the congested road section based on the drone’s field of view mapping of the ground under congestion conditions;

[0011] Step 6: Use a traffic congestion assessment method based on road section information to intelligently assess the congestion level of a traffic section based on the average speed of the section, vehicle spacing, and traffic flow density.

[0012] Furthermore, in the above step 1, the improved YOLOv5 algorithm adopts the K-MEANS clustering method to re-cluster the vehicle bounding box to reduce the false detection and missed detection caused by the large size gap between the preset anchor box and the target. On the basis of the IoU and GIoU loss functions, in order to more fully consider the bounding box center point distance information and the bounding box aspect ratio scale information, a compatible overlap rate loss function L is proposed. CIoU , to solve the problems of slow regression speed and low positioning accuracy, the formula is as follows:

[0013]

[0014] Among them, IoU represents the intersection over union ratio, A represents the predicted box, and B represents the real box. 2 (b,b gt ) represents the straight-line distance between the center points of the predicted box and the true box, b represents the diagonal distance between the minimum bounding rectangles of the two boxes, d represents the distance between the centers of the two boxes, c represents the diagonal distance between the bounding rectangles of the two boxes, a is a parameter for balancing the ratio, and v represents the aspect ratio consistency parameter.

[0015] Furthermore, in the above step 2, the calculation of the vehicle's motion state based on the drone's height, perspective, attitude, and position information includes the following: based on the drone's motion state characteristics, the position of the i-th vehicle in the drone's perspective is calculated. The formula is as follows:

[0016]

[0017] in, is the height of the drone above the ground, are the yaw angle, pitch angle, and roll angle of the drone respectively. is the angle between the UAV’s projection on the ground and the k-th vehicle, is the angle between the UAV and the k-th vehicle, is the initial position of the UAV.

[0018] Furthermore, in step 3 above, the multi-source sensor fusion technology is used to perform group intelligence evaluation on the motion information between vehicles, and the steps are as follows:

[0019] Step 3.1: Calculate the distance between the front and rear vehicles at time t1 and at time t2 after Δt1 And the speed of the i-th vehicle in this time period and the speed at which the distance between the front and rear vehicles changes

[0020] Step 3.2: Comparison and if This indicates that the relative distance between the front and rear vehicles is getting larger and larger, and there is a certain distance between the vehicles. This means that the traffic section where the i-th vehicle is located is not congested at this moment, and step 3.4 is executed.

[0021] if This indicates that the relative distance between the front and rear vehicles is getting smaller and smaller, and further judgment of the vehicle speed is required. Go to step 3.3;

[0022] Step 3.3: Comparison and if This means that during this time period, the speed of vehicle i cannot keep up with the change in the distance between the vehicles behind it, and the relative distance between the vehicles behind it is getting smaller and smaller. This means that the speed of the vehicles is slower than the degree of decrease in the distance between them. This indicates that the intersection where vehicle i is located is not congested, so proceed to step 3.4.

[0023] if This means that during this time period, the speed of the i-th vehicle has outpaced the change in the distance between the vehicles ahead and behind, and the relative distance between the vehicles behind is decreasing. This means that the speed of the vehicle is faster than the degree of decrease in the distance between the vehicles, indicating that the distance between the vehicles is decreasing. Therefore, proceed to step 3.3.

[0024] Step 3.4: Repeat steps 3.1-3.3 to calculate the distance between the front and rear vehicles of multiple vehicles in a continuous time period And the speed of each vehicle during the time period and the speed at which the distance between the front and rear vehicles changes When all vehicles before and after a certain road section meet the following conditions:

[0025]

[0026] This indicates that the distance between vehicles is decreasing during this time period, and the speeds of multiple vehicles are also decreasing. Moreover, the speed of change in the distance between vehicles is getting slower and slower. That is, the distance between vehicles is getting smaller, the speed is getting slower, and the road conditions are getting denser. In other words, based on the vehicle dynamics characteristics, it is judged that the road section is congested at this moment.

[0027] Furthermore, in the above step 4, the content of using the improved DeepSort algorithm to collect statistics on congested vehicles includes:

[0028] First, motion feature matching based on spatiotemporal correlation;

[0029] In order to achieve data association matching of motion features, the Mahalanobis distance d1(i, j) between the detection frame of the current frame and the tracking frame of the previous frame is used to calculate the distance similarity of the target position dimension. The formula is as follows:

[0030]

[0031] Where i represents the serial number of the tracking frame, j represents the serial number of the detection frame, represents the covariance matrix of the observation space predicted by the i-th NSA Kalman filter, d j Represents the observation vector of the j-th detection box, y i represents the state corresponding vector of the i-th tracking box, and T represents the transpose of the vector.

[0032] Secondly, the spatiotemporal correlation matching score s1 is introduced to take the temporal continuity and spatial consistency of the target into account.

[0033] The bounding box of the target in the previous frame is represented as A t-1 =(x t-1 ,y t-1 ,w t-1 ,h t-1 ,), the position of the current frame is A t =(x t ,y t ,w t ,h t ,), the matching score s is:

[0034]

[0035] d t =Δt

[0036] Then, based on the appearance feature matching of multi-frame comparison, the minimum cosine distance between the i-th track and the j-th track is calculated for appearance feature matching. The formula is as follows:

[0037]

[0038] Where r j represents the appearance features extracted from the j-th detection box, represents the cosine similarity of the apparent features between the i-th tracking frame and the j-th detection frame; R i Represents the appearance feature library.

[0039] Next, the multi-frame visual matching score s2 is introduced. For each target A1,…,A in the previous frame n Respectively with all targets B1,…,B in the current frame n , calculate the matching score of their feature vectors, the formula is as follows:

[0040]

[0041] Finally, the spatiotemporal correlation matching score s1 and the multi-frame visual matching score s2 are introduced at the same time, and the Mahalanobis distance and the minimum cosine distance are combined using linear weighting to achieve the overall optimal correlation matching of the data. The formula is as follows:

[0042]

[0043] Where c i,j Represents the minimum distance vector obtained by fusing Mahalanobis distance and minimum cosine distance after combining spatiotemporal correlation information and multi-frame visual matching information.

[0044] Furthermore, the calculation of the distance to the congested road section based on the drone's field of view mapping of the ground in the congested state in step 5 includes: calculating the congestion distance by fusing multi-source sensor data based on the drone's posture, viewing angle, and altitude information, and the formula is as follows:

[0045]

[0046] In t i The straight-line distance the drone moves relative to the ground at a given moment Later, I found that The traffic section in the field of view is not congested, so the calculation continues after the drone moves to this position in the air. How much mapping distance is left on the ground within the field of view Not counted, remaining ground mapping distance for:

[0047]

[0048] Furthermore, in the above step 6, the traffic congestion assessment method based on the road section information is used to conduct intelligent assessment of the traffic section congestion degree according to the average speed of the road section, the distance between vehicles and the traffic flow density, including: The calculation and evaluation are performed, and the specific formula is as follows:

[0049]

[0050] The present invention has the following beneficial effects:

[0051] 1. An improved YOLOv5 aerial vehicle detection and tracking algorithm is proposed, and vehicle bounding box detection based on improved K-MEANS clustering is performed on the aerial images taken by UAV for data preprocessing.

[0052] 2. Propose a feature extraction network based on improved EFFICIENTMET to improve detection accuracy.

[0053] 3. Calculate the vehicle's motion state based on the drone's altitude, viewing angle, posture, and position information, and use multi-source sensor fusion technology to quickly assess the congestion status of traffic sections based on the position, spacing, speed, and other motion information between vehicles.

[0054] 4. In congested situations, the system calculates the distance to congested sections based on drone field of view mapping and uses an improved DeepSort algorithm to count the number of congested sections, providing an intelligent assessment of the congestion status of traffic sections. This enables comprehensive assessment and monitoring of sudden traffic congestion, providing a more flexible and intelligent solution for traffic management and improving the efficiency of urban transportation operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0056] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments:

[0057] UAV multi-source sensing crowd intelligence assessment method for sudden traffic congestion, such as Figure 1 The specific steps are as follows:

[0058] Step 1: Use the improved YOLOv5 algorithm to detect the aerial vehicle target;

[0059] Step 2: Calculate the vehicle's motion state based on the drone's altitude, viewing angle, attitude, and position information;

[0060] Step 3: Use multi-source sensor fusion technology to perform group intelligence assessment of vehicle motion information and quickly determine whether a traffic section is congested.

[0061] Step 4: Use the improved DeepSort algorithm to count congested vehicles;

[0062] Step 5: Calculate the distance to the congested road section based on the drone’s field of view mapping of the ground under congestion conditions;

[0063] Step 6: Use a traffic congestion assessment method based on road section information to intelligently assess the congestion level of a traffic section based on the average speed of the section, vehicle spacing, and traffic flow density.

[0064] In step 1, the improved YOLOv5 algorithm uses the K-MEANS clustering method to re-cluster the vehicle bounding box to reduce the false detection and missed detection caused by the large size gap between the preset anchor box and the target. On the basis of the IoU and GIoU loss functions, in order to more fully consider the bounding box center point distance information and the bounding box aspect ratio scale information, a compatible overlap rate loss function L is proposed. CIoU , to solve the problems of slow regression speed and low positioning accuracy, the formula is as follows:

[0065]

[0066] Among them, IoU represents the intersection over union ratio, A represents the predicted box, and B represents the real box. 2 (b,b gt ) represents the straight-line distance between the center points of the predicted box and the true box, b represents the diagonal distance between the minimum bounding rectangles of the two boxes, d represents the distance between the centers of the two boxes, c represents the diagonal distance between the bounding rectangles of the two boxes, a is a parameter for balancing the ratio, and v represents the aspect ratio consistency parameter.

[0067] In step 2, the motion state of the vehicle is calculated based on the altitude, view angle, attitude, and position information of the drone, including the following: Based on the motion state characteristics of the drone, the position of the i-th vehicle in the view angle of the drone is calculated. The formula is as follows:

[0068]

[0069] in, is the height of the drone above the ground, are the yaw angle, pitch angle, and roll angle of the drone respectively. is the angle between the UAV’s projection on the ground and the k-th vehicle, is the angle between the UAV and the k-th vehicle, is the initial position of the UAV.

[0070] In step 3, multi-source sensor fusion technology is used to perform group intelligence evaluation on the motion information between vehicles. The specific steps are as follows:

[0071] Step 3.1: Calculate the distance between the front and rear vehicles at time t1 and at time t2 after Δt1 And the speed of the i-th vehicle in this time period and the speed at which the distance between the front and rear vehicles changes

[0072] Step 3.2: Comparison and if This indicates that the relative distance between the front and rear vehicles is getting larger and larger, and there is a certain distance between the vehicles. This means that the traffic section where the i-th vehicle is located is not congested at this moment, and step 3.4 is executed.

[0073] if This indicates that the relative distance between the front and rear vehicles is getting smaller and smaller, and further judgment of the vehicle speed is required. Go to step 3.3;

[0074] Step 3.3: Comparison and if This means that during this time period, the speed of the i-th vehicle cannot keep up with the speed of change in the distance between the vehicles ahead and behind, and the relative distance between the vehicles behind is getting smaller and smaller. This means that the speed of the vehicle is slower than the degree of decrease in the distance between the vehicles ahead, indicating that the traffic section where the i-th vehicle is located is not congested. Go to step 3.4.

[0075] if This means that during this time period, the speed of the i-th vehicle has outpaced the change in the distance between the vehicles ahead and behind, and the relative distance between the vehicles behind is decreasing. This means that the speed of the vehicle is faster than the degree of decrease in the distance between the vehicles, indicating that the distance between the vehicles is decreasing. Therefore, proceed to step 3.3.

[0076] Step 3.4: Repeat steps 3.1-3.3 to calculate the distance between the front and rear vehicles of multiple vehicles in a continuous time period And the speed of each vehicle during the time period and the speed at which the distance between the front and rear vehicles changes When all vehicles before and after a certain road section meet the following conditions:

[0077]

[0078] This indicates that the distance between vehicles is decreasing during this time period, and the speeds of multiple vehicles are also decreasing. Moreover, the speed of change in the distance between vehicles is getting slower and slower. That is, the distance between vehicles is getting smaller, the speed is getting slower, and the road conditions are getting denser. In other words, based on the vehicle dynamics characteristics, it is judged that the road section is congested at this moment.

[0079] The contents of the congested vehicle statistics based on the improved DeepSort algorithm in step 4 include:

[0080] First: Motion feature matching based on spatiotemporal association. In order to achieve data association matching of motion features, the Mahalanobis distance d1(i, j) between the detection frame of the current frame and the tracking frame of the previous frame is used to calculate the distance similarity of the target position dimension. The formula is as follows:

[0081]

[0082] Where i represents the serial number of the tracking frame, j represents the serial number of the detection frame, represents the covariance matrix of the observation space predicted by the i-th NSA Kalman filter, d j Represents the observation vector of the j-th detection box, y i represents the state corresponding vector of the i-th tracking box, and T represents the transpose of the vector.

[0083] Secondly, the spatiotemporal correlation matching score s1 is introduced to take into account the temporal continuity and spatial consistency of the target. The bounding box of the target in the previous frame is represented as A t-1 =(x t-1 ,y t-1 ,w t-1 ,h t-1 ,), the position of the current frame is A t =(x t ,y t ,w t ,h t ,), the matching score s is:

[0084]

[0085] d t =Δt

[0086] Then, based on the appearance feature matching of multi-frame comparison, the minimum cosine distance between the i-th track and the j-th track is calculated for appearance feature matching. The formula is as follows:

[0087]

[0088] Where r j represents the appearance features extracted from the j-th detection box, represents the cosine similarity of the apparent features between the i-th tracking frame and the j-th detection frame; R i Represents the appearance feature library.

[0089] Next, the multi-frame visual matching score s2 is introduced. For each target A1,…,A in the previous frame n Respectively with all targets B1,…,B in the current frame n , calculate the matching score of their feature vectors, the formula is as follows:

[0090]

[0091] Finally, the spatiotemporal correlation matching score s1 and the multi-frame visual matching score s2 are introduced at the same time, and the Mahalanobis distance and the minimum cosine distance are combined using linear weighting to achieve the overall optimal correlation matching of the data. The formula is as follows:

[0092]

[0093] Where c i,j Represents the minimum distance vector obtained by fusing Mahalanobis distance and minimum cosine distance after combining spatiotemporal correlation information and multi-frame visual matching information.

[0094] In step 5, the calculation of the congested road section distance based on the drone's field of view mapping of the ground in the congested state includes: calculating the congested distance by fusing multi-source sensor data based on the drone's posture, viewing angle, and altitude information. The formula is as follows:

[0095]

[0096] In t i The straight-line distance the drone moves relative to the ground at a given moment Later, I found that The traffic section in the field of view is not congested, so the calculation continues after the drone moves to this position in the air. How much mapping distance is left on the ground within the field of view Not counted, remaining ground mapping distance for:

[0097]

[0098] Step 6: Propose a traffic congestion assessment method based on road segment information, and evaluate the congestion density of traffic segments according to the average speed, vehicle spacing and traffic flow density of the road segment. The calculation and evaluation are performed, and the specific formula is as follows:

[0099]

Claims

1. A multi-source sensing crowd intelligence assessment method for sudden traffic congestion based on UAVs, characterized by: The specific steps are as follows: Step 1: Use the improved YOLOv5 algorithm to detect the aerial vehicle target; Step 2: Calculate the vehicle's motion state based on the drone's altitude, viewing angle, attitude, and position information; Step 3: Use multi-source sensor fusion technology to perform group intelligence assessment of vehicle motion information and quickly determine whether a traffic section is congested. Step 4: Use the improved DeepSort algorithm to count congested vehicles; Step 5: Calculate the distance to the congested road section based on the drone’s field of view mapping of the ground under congestion conditions; Step 6: Use a traffic congestion assessment method based on road segment information to intelligently assess the congestion level of a traffic segment based on the average speed, vehicle spacing, and traffic flow density of the road segment; The contents of the congested vehicle statistics using the improved DeepSort algorithm in step 4 include: First, based on the motion feature matching of spatiotemporal association, in order to achieve data association matching of motion features, the Mahalanobis distance d1(i, j) between the detection frame of the current frame and the tracking frame of the previous frame is used to calculate the distance similarity of the target position dimension. The formula is as follows: Where i represents the serial number of the tracking frame, j represents the serial number of the detection frame, represents the covariance matrix of the observation space predicted by the i-th NSA Kalman filter, d j Represents the observation vector of the j-th detection box, y i represents the state corresponding vector of the i-th tracking frame, T represents the transpose of the vector, Secondly, the spatiotemporal correlation matching score s1 is introduced to take into account the temporal continuity and spatial consistency of the target. The bounding box of the target in the previous frame is represented as A t-1 =(x t-1 ,y t-1 ,w t-1 ,h t-1 ,), the position of the current frame is A t =(x t ,y t ,w t ,h t ,), the matching score s is: d t =Δt Then, based on the appearance feature matching of multi-frame comparison, the minimum cosine distance between the i-th track and the j-th track is calculated for appearance feature matching. The formula is as follows: Where r j represents the appearance features extracted from the j-th detection box, represents the cosine similarity of the apparent features between the i-th tracking frame and the j-th detection frame; R i represents the apparent feature library, Next, the multi-frame visual matching score s2 is introduced. For each target A1,…,A in the previous frame n Respectively with all targets B1,…,B in the current frame n , calculate the matching score of their feature vectors, the formula is as follows: Finally, the spatiotemporal correlation matching score s1 and the multi-frame visual matching score s2 are introduced at the same time, and the Mahalanobis distance and the minimum cosine distance are combined using linear weighting to achieve the overall optimal correlation matching of the data. The formula is as follows:

2. The UAV multi-source sensing crowd intelligence assessment method for sudden traffic congestion according to claim 1 is characterized in that: The improved YOLOv5 algorithm in step 1 adopts the K-MEANS clustering method to re-cluster the vehicle bounding box to reduce the false detection and missed detection caused by the large size gap between the preset anchor box and the target. On the basis of the IoU and GIoU loss functions, in order to more fully consider the bounding box center point distance information and the bounding box aspect ratio scale information, a compatible overlap rate loss function L is proposed. CIoU , to solve the problems of slow regression speed and low positioning accuracy, the formula is as follows: Among them, IoU represents intersection over union, A represents the predicted box, B represents the real box, ρ 2 (b,b gt ) represents the straight-line distance between the center point of the predicted box and the real box, b represents the diagonal distance of the minimum circumscribed rectangle of the two boxes, d is the distance between the centers of the two boxes, c is the diagonal distance of the circumscribed rectangle of the two boxes, a is a parameter for balancing the ratio, and v represents the aspect ratio consistency parameter.

3. The UAV multi-source sensing crowd intelligence assessment method for sudden traffic congestion according to claim 2 is characterized in that: The calculation of the vehicle's motion state based on the drone's height, perspective, attitude, and position information in step 2 includes the following: Based on the drone's motion state characteristics, the position of the i-th vehicle in the drone's perspective is calculated. The formula is as follows: in, is the height of the drone above the ground, are the yaw angle, pitch angle, and roll angle of the drone respectively. is the angle between the UAV’s projection on the ground and the k-th vehicle, is the angle between the UAV and the k-th vehicle, is the initial position of the UAV.

4. The UAV multi-source sensing crowd intelligence assessment method for sudden traffic congestion according to claim 1 is characterized in that: In step 3, multi-source sensor fusion technology is used to perform group intelligence evaluation on the position, distance, speed and other motion information between vehicles. The specific steps are as follows: Step 3.1: Calculate the distance between the front and rear vehicles at time t1 and at time t2 after Δt1 And the speed of the i-th vehicle in this time period and the speed at which the distance between the front and rear vehicles changes Step 3.2: Comparison and if This indicates that the relative distance between the front and rear vehicles is getting larger and larger, and there is a certain distance between the vehicles. This means that the traffic section where the i-th vehicle is located is not congested at this moment, and step 3.4 is executed. if This indicates that the relative distance between the front and rear vehicles is getting smaller and smaller, and further judgment of the vehicle speed is required. Go to step 3.3; Step 3.3: Comparison and if This means that during this time period, the speed of the i-th vehicle cannot keep up with the speed of change in the distance between the vehicles ahead and behind, and the relative distance between the vehicles behind is getting smaller and smaller. This means that the speed of the vehicle is slower than the degree of decrease in the distance between the vehicles ahead, indicating that the traffic section where the i-th vehicle is located is not congested. Go to step 3.

4. if This means that during this time period, the speed of the i-th vehicle has outpaced the change in the distance between the vehicles ahead and behind, and the relative distance between the vehicles behind is decreasing. This means that the speed of the vehicle is faster than the degree of decrease in the distance between the vehicles, indicating that the distance between the vehicles is decreasing. Therefore, proceed to step 3.

3. Step 3.4: Repeat steps 3.1-3.3 to calculate the distance between the front and rear vehicles of multiple vehicles in a continuous time period And the speed of each vehicle during the time period and the speed at which the distance between the front and rear vehicles changes When all vehicles before and after a certain road section meet the following conditions: This indicates that the distance between vehicles is decreasing during this time period, and the speeds of multiple vehicles are also decreasing. Moreover, the speed of change in the distance between vehicles is getting slower and slower. That is, the distance between vehicles is getting smaller, the speed is getting slower, and the road conditions are getting denser. In other words, based on the vehicle dynamics characteristics, it is judged that the road section is congested at this moment.

5. The UAV multi-source sensing crowd intelligence assessment method for sudden traffic congestion according to claim 1 is characterized in that: The calculation of the congested road section distance based on the drone's field of view mapping of the ground in step 5 includes: calculating the congested distance by fusing multi-source sensor data based on the drone's posture, viewing angle, and altitude information. The formula is as follows: In t i The straight-line distance the drone moves relative to the ground at a given moment Later, I found that The traffic section in the field of view is not congested, so the calculation continues after the drone moves to this position in the air. How much mapping distance is left on the ground within the field of view Not counted, remaining ground mapping distance for:

6. The UAV multi-source sensing crowd intelligence assessment method for sudden traffic congestion according to claim 1 is characterized in that: The traffic congestion evaluation method based on the road section information is adopted in step 6, and the content of the intelligent evaluation of the traffic section congestion degree according to the average speed of the road section, the distance between vehicles and the traffic flow density includes: evaluating the traffic section congestion density according to the average speed of the road section, the distance between vehicles and the traffic flow density The calculation and evaluation are performed, and the specific formula is as follows:

Images