Vehicle-road collaborative positioning system and method

Abstract

The present invention discloses a vehicle-road collaborative positioning system and method. The vehicle-road collaborative positioning system of the present invention includes a roadside sensor unit, an edge computing unit, a roadside communication unit, a vehicle-mounted communication unit, and a vehicle-mounted computing unit that are sequentially connected for data transmission; the vehicle-mounted computing unit includes a preprocessing module, a multi-target matching tracking module, a target position prediction module, and a self-vehicle probability prediction module that are sequentially connected; the vehicle-mounted computing unit is used to locate the vehicle itself in the road area based on the roadside perception information and the vehicle's own status information. The present invention can get rid of the dependence on GNSS signals, can realize long-distance continuous positioning of networked vehicles, and can obtain the relative position relationship with surrounding vehicles, which is more practical.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent driving systems, and specifically relates to a vehicle-road collaborative positioning system and method. Background Art

[0002] In recent years, with the development of autonomous driving technology, vehicle networking communication technology, and roadside perception technology, intelligent connected vehicles and their related derivative products (such as autonomous buses, unmanned mining vehicles, unmanned delivery vehicles, etc.) have gradually become popular and put into practical use. Accurate vehicle positioning is crucial for safe and efficient vehicle driving and vehicle dispatch management. At present, the Global Navigation Satellite System (GNSS) has become the mainstream vehicle positioning technology. However, in some complex traffic scenarios (such as overpasses, tunnels, and around high-rise buildings), positioning failure is easily caused by the loss of GNSS signals.

[0003] In the prior art, for single-vehicle positioning methods in non-GNSS scenarios, positioning methods based on vision and laser odometers are easily affected by the cumulative effect, resulting in poor long-distance positioning performance; and positioning methods based on map matching have high map update and maintenance costs.

[0004] In the existing technology, there are also positioning technologies based on vehicle-to-vehicle collaboration and vehicle-to-road collaboration.

[0005] Technology based on vehicle-to-vehicle collaboration needs to rely on the vehicle mobile communication network and still relies on the GNSS signals of the vehicle itself or surrounding vehicles to assist in positioning.

[0006] Most positioning methods based on vehicle-road collaboration use signal reflection intensity and reflection time to calculate the distance and direction between the vehicle and the road communication equipment, and then realize vehicle positioning. Existing positioning methods for vehicle-road collaboration often require the deployment of a large number of roadside equipment within a short distance, and the vehicle itself cannot obtain the relative positioning relationship with other vehicles. In addition, most of the existing vehicle positioning methods based on roadside perception information remain in the simulation stage, or only rely on cameras for rough positioning, without considering the problem of collaborative positioning accuracy based on multi-source heterogeneous perception information under mixed traffic flows. In actual traffic scenarios, long-distance continuous positioning of multi-point roadside equipment and sections with different communication ranges and perception ranges is often difficult, and the difficulty and accuracy of accurately positioning the vehicle itself under mixed conditions of connected and non-connected vehicles are greatly reduced. Therefore, the existing vehicle-road collaborative positioning methods do not have good positioning accuracy and practicality. Summary of the invention

[0007] In order to overcome the defects and shortcomings of the prior art, the first purpose of the present invention is to provide a vehicle-road collaborative positioning system, and the second purpose is to provide a vehicle-road collaborative positioning method for improving the practicality and accuracy of vehicle positioning.

[0008] In order to achieve the above object, the present invention adopts the following technical solutions:

[0009] A vehicle-road collaborative positioning system comprises a roadside sensor unit, an edge computing unit, a roadside communication unit, a vehicle-mounted communication unit, and a vehicle-mounted computing unit which are sequentially connected for data transmission;

[0010] The roadside sensor unit is used to collect information about vehicles in the road area;

[0011] The edge computing unit is used to process the information collected by the roadside sensor unit to form roadside perception information of vehicles in the road area; the roadside communication unit is used to send the roadside perception information processed by the edge computing unit to the vehicle communication unit;

[0012] The vehicle-mounted communication unit is used to transmit the received roadside sensing information to the vehicle-mounted computing unit;

[0013] The vehicle-mounted computing unit is used to locate the vehicle itself in the road area based on the roadside perception information and the vehicle's own state information;

[0014] The vehicle-mounted computing unit includes a pre-processing module, a multi-target matching and tracking module, a target position prediction module, and an ego-vehicle probability prediction module which are connected in sequence;

[0015] The preprocessing module is connected to the vehicle communication unit and is used to preprocess the roadside sensing information and filter out invalid non-vehicle targets and repeated vehicle targets;

[0016] The multi-target matching and tracking module is used to obtain and track the position and speed changes of vehicle targets based on roadside perception information;

[0017] The target position prediction module is used to predict the roadside perception information and correct the position of the vehicle target according to the position and speed changes of the vehicle target using the extended Kalman filter, and then update the roadside perception information;

[0018] The ego-vehicle probability prediction module is used to calculate the probability that the currently tracked vehicle target is the vehicle itself based on the updated roadside perception information, thereby realizing the positioning of the vehicle.

[0019] Preferably, the roadside sensor unit includes a plurality of sensors respectively deployed at a plurality of points in the road area, and the types of the sensors include laser radar and camera;

[0020] LiDAR and cameras are used to obtain point cloud and image data of road areas;

[0021] Each sensor is electrically connected to an edge computing unit for data transmission, and each sensor is provided with global positioning coordinates and joint calibration parameters corresponding to the current coordinate system of the road area.

[0022] Furthermore, the edge computing unit includes a target detection module, a perception fusion module, a coordinate conversion module, and an encoding module;

[0023] The target detection module is connected to multiple sensors in the roadside sensor unit, and is used to perform two-dimensional target detection on the images and point cloud data collected by the roadside multi-sensors and three-dimensional target detection on the points, and output target detection information of different sensors; the target detection information includes the geometric size, position, and detection frame of the vehicle target;

[0024] The perception fusion module is connected to the target detection module to perform data matching, association, and misdetection elimination processing on the image two-dimensional target detection information and the point cloud three-dimensional target detection information, and output the processed three-dimensional target detection information as roadside perception information;

[0025] The coordinate conversion module is connected to the perception fusion module and is used to convert the coordinate points in the current coordinate system of the road area into the longitude and latitude coordinates in the world coordinate system;

[0026] The encoding module is connected to the coordinate conversion module and is used to encode the roadside perception information according to the standard C-V2X data communication protocol.

[0027] Furthermore, the preprocessing module is used to receive the encoded roadside perception information, decode the roadside perception information, convert the longitude and latitude coordinates of the world coordinate system in the roadside perception information into coordinates in the current road area coordinate system, and eliminate invalid data such as non-vehicle targets and repeated detection targets based on the roadside perception information.

[0028] Furthermore, the process of obtaining and tracking the position and speed changes of multiple vehicle targets is as follows:

[0029] Traverse the roadside perception information of all detected vehicle targets and all tracked vehicle targets in the current frame, and set them as detection targets and tracking targets respectively;

[0030] If the current frame is the first frame to start tracking, each detected target in the current frame will be directly used as the tracking target of the next frame;

[0031] If the current frame is not the first frame to start tracking, the geometric size intersection and union ratio, position distance, and speed derived from the position distance and the time difference between the previous and next frames of each detected target in the current frame and all tracked targets are calculated;

[0032] The geometric size intersection-union ratio is calculated as follows: the geometric center points of the detection frames of the detected target and the tracked target are overlapped, and the ratio of the intersection volume to the union volume of the two detection frames is calculated;

[0033] The position distance is calculated by calculating the Euclidean distance between the geometric center points of the detection frame of the detected target and the tracked target;

[0034] The speed is calculated by dividing the distance between the detection frame of the detected target and the tracked target by the time interval between the previous and next frames;

[0035] Set the speed matching probability Prob vel And the calculation is shown as follows:

[0036]

[0037] Among them, v c 、v t 、v thresh Respectively represent the speed of the detected target in the current frame, the speed of the tracked target, and the set speed threshold;

[0038] Set the distance matching probability Prob dis And the calculation is shown as follows:

[0039]

[0040] Among them, d o ,d thresh Respectively represent the distance between the detected target and the tracked target in the current frame and the set distance threshold;

[0041] Set the geometric size intersection ratio to the IOU matching probability Prob iou ;

[0042] According to the IOU matching probability Prob iou , distance matching probability Prob dis , speed matching probability Prob vel , calculate the target matching probability Prob of the detected target and all tracked targets in the current frame m , target matching probability Prob m As shown below:

[0043] Prob m =w iou Prob iou +w vel Prob vel +w dis Prob dis

[0044] Among them, w iou 、w vel 、w dis Corresponding to IOU matching probability Prob iou , speed matching probability Prob vel, distance matching probability Prob dis their respective weight values;

[0045] After all the detection targets in the current frame have been traversed accordingly, according to the target matching probability Prob m Update the roadside perception information of all tracked targets.

[0046] Further, according to the target matching probability Prob m The process of updating the roadside perception information of all tracked targets is as follows:

[0047] If there is a detection target corresponding to part or all of the tracking targets, the target matching probability Prob m If it is greater than the set matching threshold, the target matching probability Prob is judged m The highest corresponding tracking target and the detection target in the current frame are the same vehicle, and then the roadside perception information of the detection target in the current frame is used to replace the roadside perception information of the corresponding tracking target;

[0048] If there is a detection target in the current frame that corresponds to all tracked targets, the target matching probability Prob m If the value is less than the set matching threshold, it is determined that the detected target cannot find the corresponding tracking target in the previous frame, and then the roadside perception information of the detected target is directly used as the roadside perception information of the corresponding tracking target in the next frame.

[0049] Furthermore, the process of using the extended Kalman filter to predict the roadside perception information and correct the target position is as follows:

[0050] If there is a detected target in the current frame that cannot be matched with the tracked target, an extended Kalman filter is constructed and initialized using the position of the detected target in the current frame;

[0051] If there is a detection target in the current frame that matches a certain tracking target, the extended Kalman filter corresponding to the tracking target is predicted using the position and speed corresponding to the detection target in the current frame, and the corrected position is output as the latest position of the detection target;

[0052] If there is a tracking target that cannot match the detection target of the current frame, the roadside perception information of the tracking target and its extended Kalman filter are used, combined with the time interval between the current frame and the previous frame, to predict and update the position of the tracking target corresponding to the current frame.

[0053] Furthermore, the process of calculating the probability that the vehicle target belongs to the vehicle itself is as follows:

[0054] Traverse the roadside perception information of all vehicle targets in the current frame after the update, and statistically calculate the geometric size, position, speed, and moved position distance of each vehicle target;

[0055] According to the geometric size of the vehicle itself, the speed of the vehicle itself, and the position and distance information of the vehicle itself, the latest intersection and union ratio of the detection frame of the vehicle itself and all vehicle targets is calculated respectively. Average Intersection-Union Ratio The probability that the vehicle target belongs to the vehicle itself Probability of the vehicle moving distance Then, based on the adaptive weighted method of coefficient of variation statistics, the weighted self-vehicle matching probability Prob that the vehicle target belongs to the vehicle itself is calculated ego , according to the weighted self-car matching probability Prob ego Determine whether the vehicle target corresponding to the corresponding tracking sequence belongs to the vehicle itself;

[0056] Vehicle speed probability Probability of the vehicle moving distance Weighted vehicle matching probability Prob ego The calculation formula is as follows:

[0057]

[0058]

[0059]

[0060] Among them, v t 、v e Respectively represent the speed of the vehicle target and the speed of the vehicle itself, d t ,d e ,d thresh They represent the moving distance of the vehicle target, the moving distance of the vehicle, and the moving distance threshold, respectively. i represents the weight, Represents the weighted item corresponding to type i.

[0061] Furthermore, the adaptive weighting method based on coefficient of variation statistics includes:

[0062] Compare the latest intersection Average Intersection-Union Ratio Vehicle speed probability Probability of the vehicle moving distance After the weights are initialized, the intersection-over-combination ratio variation coefficient and the speed variation coefficient are set to perform adaptive adjustment on different weights. The calculation of adaptive adjustment is shown in the following formula:

[0063]

[0064]

[0065] Among them, CV j represents the coefficient of variation of the intersection-and-joint ratio of a vehicle target history or the coefficient of variation of the speed of the history, j represents the corresponding intersection-and-joint ratio or speed, Indicates the maximum value of the intersection-over-combination coefficient of variation of all vehicle target histories or the speed coefficient of variation of the history in the current frame, A iou , A vel They represent the adaptive coefficient of the intersection-and-joint ratio and the adaptive coefficient of the speed of a certain vehicle target history, respectively. Represents the initialization weights.

[0066] A vehicle-road collaborative positioning method comprises the following steps:

[0067] Collect information about vehicles in the road area;

[0068] Processing the collected information to form roadside perception information of vehicles in the road area;

[0069] Pre-process the roadside perception information of each vehicle to filter out invalid non-vehicle targets and repeated vehicle targets;

[0070] Based on the roadside perception information of multiple vehicles, the position and speed changes of vehicle targets are obtained and tracked;

[0071] According to the position and speed changes of the vehicle target, the extended Kalman filter is used to predict and correct the position of the roadside perception information set, and then update the roadside perception information;

[0072] The probability that the currently tracked vehicle target is the vehicle itself is calculated based on the updated roadside perception information, thereby achieving the positioning of the vehicle.

[0073] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0074] The vehicle-road collaborative positioning system and method of the present invention relies on roadside perception information from various types of sensor detection sources, the geometric dimensions and speed information of the vehicle itself to perform vehicle-road collaborative positioning. Compared with the existing positioning method based on vehicle-to-vehicle collaboration, it can get rid of the dependence on GNSS signals. Compared with the existing vehicle-road collaborative positioning method based on the roadside RSU receiving signal strength, it can realize long-distance continuous positioning of networked vehicles, and can obtain the relative position relationship with surrounding vehicles, which is more practical.

[0075] Furthermore, the present invention can predict the location information of the vehicle of the current tracked target based on historical data when the roadside perception information is lost by predicting the target location. Compared with a single GPS positioning or a single roadside perception positioning method, the present invention can effectively increase the positioning signal output frequency and provide more real-time positioning information for the connected vehicle.

[0076] Furthermore, the present invention, through the prediction of the probability of the own vehicle and the adaptive weighting method based on the statistics of the coefficient of variation, can effectively locate the vehicle's own position information when connected vehicles and non-connected vehicles are mixed. It has high positioning accuracy and positioning precision and has practical application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 It is a schematic diagram of the structural framework of the vehicle-road collaborative positioning system of the present invention;

[0078] Figure 2 It is a schematic diagram of the process flow of the vehicle-road collaborative positioning method of the present invention;

[0079] Figure 3 for Figure 2 Flowchart of multi-target matching and tracking in ;

[0080] Figure 4 for Figure 2 Flowchart for target position prediction in ;

[0081] Figure 5 for Figure 2 Flowchart for ego-vehicle probability prediction in . DETAILED DESCRIPTION

[0082] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0083] It should be noted that the orientations or positional relationships indicated by terms such as “center”, “up”, “down”, “left”, “right”, “vertical”, “horizontal”, “inside” and “outside” are based on the orientations or positional relationships shown in the accompanying drawings and are only for the convenience of describing the present disclosure and simplifying the description. They do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present disclosure.

[0084] In addition, the terms "first", "second", and "third" are used for descriptive purposes only and are not to be understood as indicating or implying relative importance. Similarly, words such as "one", "an", or "the" do not indicate a quantitative limitation, but rather indicate the presence of at least one. Words such as "include" or "comprise" and the like mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. Words such as "electrically connected" or "connected" and the like are not limited to physical or mechanical electrical connections, but may include electrical connections, whether direct or indirect.

[0085] Example 1

[0086] like Figure 1 This embodiment 1 provides a vehicle-road collaborative positioning system, which is particularly suitable for vehicle-road collaborative positioning in traffic scenarios such as parks, urban road networks, and highways, and is particularly suitable for vehicle-road collaborative positioning in scenarios with multi-sensing, multi-point, and multi-view roadside perception.

[0087] The system of this embodiment includes a roadside sensor unit, an edge computing unit (MEC), a roadside communication unit (RSU), an onboard communication unit (OBU), and an onboard computing unit, which are connected in sequence and can perform bidirectional data transmission. In this embodiment, the onboard communication unit (OBU) and the onboard computing unit are preferably integrated into the vehicle system of the intelligent connected vehicle.

[0088] The roadside sensor unit is arranged in the road area, and multiple roadside sensor units can be deployed at multiple points in the same continuous road area. Each roadside sensor unit is provided with one or more types of sensors, including lidar, camera, millimeter wave radar, etc. Each type of sensor has a different perception range and perception angle. The combination of multiple types of sensors can cover the information fusion perception range of the road area as much as possible to achieve the function of multi-perspective. Each sensor is electrically connected to an edge computing unit for data transmission, and each sensor is provided with global positioning coordinates and joint calibration parameters corresponding to the current coordinate system of the road area. The roadside sensor unit is used for information about vehicles in the road area, and the data form of the information includes images, point cloud data, etc. In this embodiment, preferably, the roadside sensor unit is installed on an intelligent road pole, and each sensor transmits data with the edge computing unit through direct communication via a network cable.

[0089] The edge computing unit is arranged in the road area. In the same continuous road area, multiple edge computing units can be deployed at multiple points respectively, or an edge computing unit at one point can be connected to roadside sensor units at multiple points. In the preferred embodiment of the present invention, the specific type of the edge computing unit is a dedicated computing device, which is also installed on the smart road pole. Various types of sensors can be mounted under each edge computing unit to realize the reception and processing of roadside perception information. The edge computing unit is used to perform joint coordinate calibration of multiple sensors in the roadside sensor unit, perform two-dimensional target detection and three-dimensional target detection based on images and point cloud data to obtain target detection information, convert roadside perception information in different sensor coordinate systems into a unified world coordinate system, and encode the three-dimensional target detection data after the coordinate system is converted.

[0090] The edge computing unit includes a target detection module, a perception fusion module, a coordinate conversion module, and an encoding module.

[0091] The target detection module is connected to multiple sensors in the roadside sensor unit, and is used to perform two-dimensional target detection on the images and point cloud data collected by the roadside multi-sensors, and output target detection information. The target detection information includes the category, geometric size, center point coordinates, etc. of the vehicle target.

[0092] The perception fusion module is connected to the target detection module, and is used to match and associate the image 2D target detection information with the point cloud 3D target detection information, eliminate false detections, and output the fused 3D target detection information as roadside perception information. The 3D target detection information includes the 3D target detection box and detection box information in the point cloud.

[0093] The coordinate conversion module is connected to the perception fusion module for data transmission. The coordinate conversion module is used to uniformly convert the coordinate points of various data such as sensors, vehicle targets, three-dimensional target detection frames, etc. in the current coordinate system of the road area into longitude and latitude coordinates in the world coordinate system, and then combine the longitude and latitude coordinates with the information of the detection frame to form roadside perception information (RSM). The specific data form of the roadside perception information includes detection frames, target detection information, and longitude and latitude coordinates. In this embodiment, the coordinate conversion of each sensor and each vehicle is based on the coordinates of its center point. In this embodiment, it is further preferred that the roadside perception information generated by the perception fusion module also carries timestamp data. In this embodiment, it is further preferred that the roadside perception information of each vehicle target is recorded in the form of a data sequence, and the data sequence is named a tracking sequence. The tracking sequences of all vehicle targets constitute a roadside perception information set.

[0094] The encoding module is connected to the coordinate conversion module for data transmission. The encoding module is used to encode the roadside perception information according to the set data communication protocol to form a data format suitable for wireless communication transmission. In this embodiment, the encoding module preferably encodes the roadside perception information according to the C-V2X protocol.

[0095] The roadside communication unit is electrically connected to the encoding module for data transmission. The roadside communication unit is used to wirelessly transmit the roadside sensing information after data encoding, so that vehicles in the road area can wirelessly receive the corresponding roadside sensing information, completing the wireless communication transmission of the roadside sensing information.

[0096] The vehicle-mounted communication unit and the roadside communication unit perform data transmission in a wireless communication manner. In this embodiment, the vehicle-mounted communication unit and the roadside communication unit preferably use the C-V2X protocol for wireless communication and send it in a broadcasting manner. The vehicle-mounted communication unit is used to receive the corresponding roadside perception information.

[0097] The on-board computing unit is used to obtain and store the vehicle's own speed data, geometric dimension data, and vehicle category information.

[0098] The vehicle-mounted computing unit is electrically connected to the vehicle-mounted communication unit for data transmission, and is used to locate the corresponding vehicle target in the road area. The vehicle-mounted computing unit includes a preprocessing module, a multi-target matching and tracking module, a target position prediction module, and a self-vehicle probability prediction module.

[0099] The preprocessing module is connected to the vehicle communication unit. The preprocessing module is used to process the data required to locate the corresponding vehicle, receive the encoded roadside perception information, and then decode the roadside perception information, convert the longitude and latitude coordinates of the world coordinate system in the roadside perception information into the coordinates of the current road area coordinate system, and then filter and remove invalid data such as non-vehicle targets and repeated detection targets according to the vehicle category, timestamp and other information in the roadside perception information, and then synchronize the timestamp of the collected roadside perception information. Timestamp synchronization is used to determine the detection frame, target detection information, coordinates and other information corresponding to each frame at the corresponding time point.

[0100] The multi-target matching and tracking module is connected with the pre-processing module for data transmission. The multi-target matching and tracking module is used to obtain and track the position and speed changes of multiple vehicle targets in the road area.

[0101] The method by which the multi-target matching and tracking module acquires and tracks vehicle targets in the road area is as follows:

[0102] Traverse the tracking root sequences of the roadside sensing information corresponding to all detected vehicle targets and all tracked vehicle targets in the current frame, and set them as detection targets and tracking targets respectively;

[0103] If the current frame is the first frame to start tracking, each detected target in the current frame is directly used as the tracking target of the next frame, and the relevant data in the tracking sequence of the detected target is used as the tracking sequence related data of the tracking target of the next frame;

[0104] If the current frame is not the first frame to start tracking, the geometric size intersection over union (IOU), position distance, and speed derived from the position distance and the time difference between the previous and next frames of each detected target in the current frame and all tracked targets are calculated from the tracking sequence of the detected target and the tracking sequence of the tracked target;

[0105] The geometric size intersection-union ratio is calculated as follows: the geometric center points of the three-dimensional detection frames of the detected target and the tracked target are overlapped, and the ratio of the intersection volume to the union volume of the two three-dimensional detection frames is calculated;

[0106] The position distance is calculated by calculating the Euclidean distance between the geometric center points of the three-dimensional detection boxes of the detected target and the tracked target;

[0107] The speed is calculated by dividing the distance between the three-dimensional detection frame of the detected target and the tracked target by the time interval between the previous and next frames;

[0108] Filter out the tracking target corresponding to the previous frame for the detected target, realize the function of matching the detected target with the tracking target in the current frame, and introduce the IOU matching probability Prob iou , distance matching probability Prob dis , speed matching probability Prob vel To help with screening;

[0109] Set the speed matching probability Prob vel The calculation of is as follows:

[0110]

[0111] Among them, v c 、v t 、v thresh They represent the speed of the detected target in the current frame, the speed of the tracked target, and the set speed threshold. In this embodiment, the preferred speed threshold is set to v thresh =4m / s;

[0112] Set the distance matching probability Prob dis The calculation of is as follows:

[0113]

[0114] Among them, d o ,d threshd represents the distance between the detected target and the tracked target in the current frame and the set distance threshold. In the embodiment, the distance threshold is preferably set to d thresh =20m, △t is the time interval between the current frame and the previous frame;

[0115] Set the geometric size intersection-over-union (IOU) value itself as the IOU matching probability Prob iou ;

[0116] According to the IOU matching probability Prob iou , distance matching probability Prob dis , speed matching probability Prob vel , calculate the target matching probability Prob of the detected target in the current frame and all tracked targets in the previous frame m , target matching probability Prob m The calculation is as follows:

[0117] Prob m =w iou Prob iou +w vel Prob vel +w dis Prob dis

[0118] Among them, w iou 、w vel 、w dis Corresponding to IOU matching probability Prob iou , speed matching probability Prob vel , distance matching probability Prob dis The weight values ​​of each, in this embodiment, the preferred IOU matching probability Prob iou The weight value is set to w iou =0.2, speed matching probability Prob vel The weight value is set to w vel =0.3, distance matching probability Prob dis The weight value is set to w dis =0.5;

[0119] After all the detection targets of the current frame have been traversed accordingly, according to the target matching probability Prob m , update the roadside perception information of all detected targets according to whether there are matching tracking targets; if there is a detection target corresponding to some or all tracking targets, the target matching probability Prob m If it is greater than the set matching threshold, the target matching probability Prob is judged mThe highest corresponding tracking target is the same vehicle as the detection target in the current frame, and then the tracking sequence of the detection target in the current frame is replaced with the tracking sequence of the detection target to form a new roadside perception information set, thereby preserving the geometric size, position, speed and other information of the detection target, so that the roadside perception information of the detection target is updated to the roadside perception information of the tracking target corresponding to the next frame; if there is a detection target in the current frame corresponding to all the tracking targets, the target matching probability Prob m If the detection target is smaller than the set matching threshold, it is determined that the detection target cannot find the corresponding tracking target in the previous frame, and then a new tracking sequence is established for the detection target in the current frame, and then the new tracking sequence is added to the historical roadside perception information set, so as to save the geometric size and position information of the detection target in the current frame, so that the roadside perception information of the detection target becomes the roadside perception information of a newly added tracking target in the next frame; the preferred target matching probability Prob in this embodiment is m The corresponding matching threshold is set to 0.5.

[0120] The target position prediction module is connected to the multi-target matching and tracking module for data transmission. The target position prediction module uses an extended Kalman filter to predict the roadside perception information and correct the position of the vehicle target according to the position and speed changes of the vehicle target, and then updates the roadside perception information.

[0121] According to the results of multi-target matching and tracking in the previous step, the method of using the extended Kalman filter to predict and correct the position of the vehicle target is as follows:

[0122] If there is a detected target in the current frame that cannot be matched with the tracked target, an extended Kalman filter is constructed and initialized using the position of the detected target in the current frame;

[0123] If there is a detection target in the current frame that matches a certain tracking target, the extended Kalman filter corresponding to the tracking target is predicted using the position and speed corresponding to the detection target in the current frame, and the corrected position is output as the latest position of the detection target;

[0124] If there is a tracking target that cannot match the detection target of the current frame, the roadside perception information of the tracking target and its extended Kalman filter are used, combined with the time interval between the current frame and the previous frame, to predict and update the position of the tracking target corresponding to the current frame;

[0125] After the positions of all vehicle targets are updated in the above three situations, the new roadside perception information set of the current frame is also updated accordingly.

[0126] The ego vehicle probability prediction module is connected to the target position prediction module for data transmission. The ego vehicle probability prediction module is used to calculate the probability that the vehicle target belongs to the vehicle itself based on the roadside perception information set updated by the target position prediction module. The method for calculating the probability that the vehicle target belongs to the vehicle itself is as follows:

[0127] Receive the roadside perception information set updated by the target position prediction module, traverse all tracking sequences in the updated roadside perception information set, and count the geometric size, target position, target speed, and moved position distance of the vehicle target in each tracking sequence;

[0128] According to the geometric size, speed and moving distance of the vehicle, the latest intersection and union ratio between the vehicle and the three-dimensional detection frames of all vehicle targets is calculated. Average Intersection-Union Ratio The probability that the vehicle target belongs to the vehicle itself Probability of the vehicle moving distance Then, the weighted self-vehicle matching probability Probego of the vehicle target belonging to the vehicle itself is calculated, and the vehicle target corresponding to the corresponding tracking sequence is judged whether it belongs to the vehicle itself according to the weighted self-vehicle matching probability Probego;

[0129] The relevant calculation formula is shown as follows:

[0130]

[0131]

[0132]

[0133] Among them, v t 、v e Respectively represent the speed of the vehicle target and the speed of the vehicle itself, d t ,d e ,d thresh They represent the moving distance of the vehicle target, the moving distance of the vehicle, and the moving distance threshold, respectively. i represents the weight, represents the weighted item corresponding to type i;

[0134] In order to better distinguish the vehicle itself from the surrounding vehicles, this embodiment preferably further adopts an adaptive weighting method based on the statistics of the coefficient of variation to Average Intersection-Union Ratio Vehicle speed probability Probability of the vehicle moving distance After the weights are initialized, the intersection-over-combination ratio variation coefficient and the speed variation coefficient are set to perform adaptive adjustment on different weights. The calculation of adaptive adjustment is shown in the following formula:

[0135]

[0136]

[0137] Among them, CV j represents the coefficient of variation of the intersection-and-joint ratio of a vehicle target history or the coefficient of variation of the speed of the history, j represents the corresponding intersection-and-joint ratio or speed, Indicates the maximum value of the intersection-over-combination coefficient of variation of all vehicle target histories or the speed coefficient of variation of the history in the current frame, A iou , A vel They represent the adaptive coefficient of the intersection-and-joint ratio and the adaptive coefficient of the speed of a certain vehicle target history, respectively. i init represents the initialization weights. In this embodiment, the initialization weights are preferably set to be

[0138] The adaptively adjusted weight w i Substitute the weighted vehicle matching probability Prob ego The calculation formula is used to obtain the weighted self-vehicle matching probability Prob of the corresponding tracking sequence belonging to the vehicle itself after adaptive weighted calculation. ego , take the weighted self-vehicle matching probability Prob ego The target position of the highest vehicle target is used as the current position of the vehicle itself.

[0139] During the operation of this embodiment, the data information transmission chain is as follows: from the roadside sensor unit as the data source, it passes through the perception fusion module, the target detection module, the coordinate conversion module, the encoding module, the roadside communication unit, the vehicle communication unit, the preprocessing module, the multi-target matching and tracking module, the target position prediction module, and finally to the vehicle probability prediction module to obtain the final vehicle positioning information.

[0140] Compared with the prior art, the beneficial effects of this embodiment 1 are:

[0141] This embodiment relies on roadside perception information from various types of sensor detection sources, the vehicle's own geometric dimensions and speed information to perform vehicle-road collaborative positioning. Compared with the existing positioning method based on vehicle-to-vehicle collaboration, this embodiment can get rid of the dependence on GNSS signals. Compared with the existing vehicle-to-road collaborative positioning method based on the roadside RSU receiving signal strength, this embodiment can achieve long-distance continuous positioning of networked vehicles and obtain the relative position relationship with surrounding vehicles, which is more practical.

[0142] Furthermore, this embodiment uses a target location prediction module to predict the location information of the vehicle being tracked based on historical data when the roadside sensing information is lost. Compared with a single GPS positioning or a single roadside sensing positioning method, it can effectively increase the positioning signal output frequency and provide more real-time positioning information for the connected vehicle.

[0143] Furthermore, this embodiment uses the self-vehicle probability prediction module, combined with an adaptive weighted method based on coefficient of variation statistics, to effectively locate the vehicle's own position information when connected vehicles and non-connected vehicles are mixed. It has high positioning accuracy and positioning precision and has practical application value.

[0144] Example 2

[0145] like Figures 2 to 5 As shown. This embodiment 2 provides a vehicle-road collaborative positioning method. This embodiment preferably performs this method based on the vehicle-road collaborative positioning system in embodiment 1. The steps of the vehicle-road collaborative positioning method in this embodiment 2 are as follows:

[0146] S1. Use roadside cameras, laser radar and other sensors to collect images and point cloud data in the road area, perform coordinate conversion after vehicle target recognition, and then encode the data. The roadside perception data with target detection information and world coordinate system is wirelessly transmitted with vehicles in the road area through roadside communication equipment;

[0147] The specific process includes:

[0148] S11, using a roadside sensor unit to collect point cloud and image data of vehicles in the same continuous road area;

[0149] In this embodiment, the sensor is preferably provided with global positioning coordinates and joint calibration parameters corresponding to the current coordinate system of the road area;

[0150] S12, using the roadside sensor unit to transmit the image and point cloud data of the road environment to the target detection module in the edge computing unit, performing image two-dimensional target detection and point cloud three-dimensional target detection on the image and point cloud data collected by the roadside multi-sensor, and outputting the perception information of different sensors;

[0151] S13, using the perception fusion module to perform data matching and association between the image two-dimensional object detection information and the point cloud three-dimensional object detection information, eliminate false detection processing, and output the three-dimensional object detection information after fusion processing;

[0152] In this embodiment, the target detection information includes the type, geometric size, center point coordinates, and detection frame of the vehicle, and the detection frame includes a two-dimensional detection frame and a three-dimensional detection frame;

[0153] S14, using a coordinate conversion module to uniformly convert the coordinate points of the sensor and the vehicle target in the current coordinate system of the road area into longitude and latitude coordinates in the world coordinate system on the three-dimensional target detection information after the perception information fusion, and then combine the longitude and latitude coordinates with the target detection information to form roadside perception information;

[0154] S15, using the encoding module to encode the roadside sensing information according to the C-V2X protocol to form a data format suitable for wireless communication transmission;

[0155] S16, using the roadside communication unit to send the data-encoded roadside sensing information to the respective on-board communication units of each intelligent connected vehicle in the road area in a wireless communication manner according to the C-V2X protocol;

[0156] S2, when the on-board communication unit of the intelligent connected vehicle receives the roadside sensing information, it uses the vehicle status unit to obtain and store the vehicle's own speed data, geometric size data, and vehicle category, and then the on-board communication unit and the vehicle status unit transmit the roadside sensing information, the vehicle's own data, and the vehicle category to the on-board computing unit respectively;

[0157] S3. The intelligent connected vehicle uses the preprocessing module of the on-board computing unit to receive the encoded roadside perception information, decode the roadside perception information according to the C-V2X protocol, and then convert the longitude and latitude coordinates of the world coordinate system in the roadside perception information into the coordinates of the current road area coordinate system. Then, according to the vehicle category, timestamp and other information in the roadside perception information, invalid data such as non-vehicle targets and repeated detection targets are filtered and removed, and then the collected roadside perception information is synchronized with the timestamp.

[0158] In this embodiment, preferably, the timestamp synchronization is used to determine the detection frame, target detection information, coordinates, vehicle category, vehicle speed, and vehicle geometric size information corresponding to each frame at the corresponding time point;

[0159] S4, using a multi-target matching and tracking module of the vehicle-mounted computing unit to obtain and track the position and speed changes of multiple vehicle targets in the road area;

[0160] The specific process of multi-target matching and tracking includes:

[0161] S41, traversing the tracking root sequences of the roadside sensing information corresponding to all detected vehicle targets and all tracked vehicle targets in the current frame, and setting them as detection targets and tracking targets respectively;

[0162] S42, if the current frame is the first frame to start tracking, then each detected target in the current frame is directly used as the tracking target of the next frame, and the relevant data in the tracking sequence of the detected target is used as the tracking sequence related data of the tracking target of the next frame;

[0163] S43, if the current frame is not the first frame to start tracking, then from the tracking sequence of the detected target and the tracking sequence of the tracked target, calculate the geometric size intersection over union (IOU), position distance, and speed derived from the position distance and the time difference between the previous and next frames of each detected target in the current frame and all tracked targets;

[0164] The geometric size intersection-union ratio is calculated as follows: the geometric center points of the three-dimensional detection frames of the detected target and the tracked target are overlapped, and the ratio of the intersection volume to the union volume of the two three-dimensional detection frames is calculated;

[0165] The position distance is calculated by calculating the Euclidean distance between the geometric center points of the three-dimensional detection boxes of the detected target and the tracked target;

[0166] The speed is calculated by dividing the distance between the three-dimensional detection frame of the detected target and the tracked target by the time interval between the previous and next frames;

[0167] Filter out the tracking target corresponding to the previous frame for the detected target, realize the function of matching the detected target with the tracking target in the current frame, and introduce the IOU matching probability Prob iou , distance matching probability Prob dis , speed matching probability Prob vel To help with screening;

[0168] Set the speed matching probability Prob vel The calculation of is as follows:

[0169]

[0170] Among them, v c 、v t 、v thresh They represent the speed of the detected target in the current frame, the speed of the tracked target, and the set speed threshold. In this embodiment, the preferred speed threshold is set to v thresh =4m / s;

[0171] Set the distance matching probability Prob dis The calculation of is as follows:

[0172]

[0173] Among them, d o ,d thresh d represents the distance between the detected target and the tracked target in the current frame and the set distance threshold. In the embodiment, the distance threshold is preferably set to d thresh =20m, △t is the time interval between the current frame and the previous frame;

[0174] Set the geometric size intersection-over-union (IOU) value itself as the IOU matching probability Prob iou ;

[0175] S44, according to IOU matching probability Prob iou , distance matching probability Prob dis , speed matching probability Prob vel , calculate the target matching probability Prob of the detected target in the current frame and all tracked targets in the previous frame m , target matching probability Prob m The calculation is as follows:

[0176] Prob m =w iou Prob iou +w vel Prob vel +w dis Prob dis

[0177] Among them, w iou 、w vel 、w dis Corresponding to IOU matching probability Prob iou , speed matching probability Prob vel , distance matching probability Prob dis The weight values ​​of each, in this embodiment, the preferred IOU matching probability Prob iou The weight value is set to w iou =0.2, speed matching probability Prob vel The weight value is set to w vel =0.3, distance matching probability Prob dis The weight value is set to w dis =0.5;

[0178] S45, after all the detection targets of the current frame have been traversed accordingly, according to the target matching probability Prob m , update the roadside perception information of all detected targets according to whether there is a matching tracking target;

[0179] S451, if there is a detection target corresponding to part or all of the tracking targets, the target matching probability Prob m If it is greater than the set matching threshold, the target matching probability Prob is judged mThe highest corresponding tracking target is the same vehicle as the detection target in the current frame, and then the tracking sequence of the detection target in the current frame is replaced with the tracking sequence of the detection target to form a new roadside perception information set, thereby preserving the geometric size, position, speed and other information of the detection target, so that the roadside perception information of the detection target is updated to the roadside perception information of the tracking target corresponding to the next frame;

[0180] S452: If there is a detection target in the current frame that corresponds to all tracked targets, the target matching probability Prob m If the detection target is smaller than the set matching threshold, it is determined that the detection target cannot find the corresponding tracking target in the previous frame, and then a new tracking sequence is established for the detection target in the current frame, and then the new tracking sequence is added to the historical roadside perception information set, so as to save the geometric size and position information of the detection target in the current frame, so that the roadside perception information of the detection target becomes the roadside perception information of a newly added tracking target in the next frame; the preferred target matching probability Prob in this embodiment is m The corresponding matching threshold is set to 0.5;

[0181] S5, using the target position prediction module, using the extended Kalman filter to predict and correct the position of the vehicle target in the current frame of the roadside perception information set, and updating the roadside perception information set;

[0182] The specific process includes:

[0183] S51, if there is a detected target in the current frame that cannot be matched with the tracked target, construct and initialize an extended Kalman filter using the position of the detected target in the current frame;

[0184] S52, if there is a detection target in the current frame that can match a certain tracking target, use the position and speed corresponding to the detection target in the current frame to predict the extended Kalman filter corresponding to the tracking target, and output the corrected position as the latest position of the detection target;

[0185] S53, if there is a tracking target that cannot match the detection target of the current frame, using the roadside perception information of the tracking target and its extended Kalman filter, combined with the time interval between the current frame and the previous frame, to predict and update the position of the tracking target corresponding to the current frame;

[0186] S6, using the vehicle probability prediction module to calculate the probability that the vehicle target belongs to the vehicle itself;

[0187] The specific process includes:

[0188] S61, receiving a roadside sensing information set updated by a target position prediction module, traversing all tracking sequences in the updated roadside sensing information set, and counting geometric dimensions, target positions, target speeds, and moved position distances of vehicle targets in each tracking sequence;

[0189] S62, according to the geometric size of the vehicle itself, the speed of the vehicle itself, and the position distance information of the vehicle itself, respectively calculate the latest intersection and union ratio between the vehicle itself and the three-dimensional detection frames of all vehicle targets. Average Intersection-Union Ratio The probability that the vehicle target belongs to the vehicle itself Probability of the vehicle moving distance Then, the weighted self-vehicle matching probability Probego of the vehicle target belonging to the vehicle itself is calculated, and the vehicle target corresponding to the corresponding tracking sequence is judged whether it belongs to the vehicle itself according to the weighted self-vehicle matching probability Probego;

[0190] Vehicle speed probability Probability of the vehicle moving distance The calculation formula of weighted self-vehicle matching probability Probego is as follows:

[0191]

[0192]

[0193]

[0194] Among them, v t 、v e Respectively represent the speed of the vehicle target and the speed of the vehicle itself, d t ,d e ,d thresh They represent the moving distance of the vehicle target, the moving distance of the vehicle, and the moving distance threshold, respectively. i represents the weight, represents the weighted item corresponding to type i;

[0195] S63, using an adaptive weighting method based on coefficient of variation statistics to Average Intersection-Union Ratio Vehicle speed probability Probability of the vehicle moving distance After the weights are initialized, the intersection-over-combination ratio variation coefficient and the speed variation coefficient are set to perform adaptive adjustment on different weights. The calculation of adaptive adjustment is shown in the following formula:

[0196]

[0197]

[0198] Among them, CV j represents the coefficient of variation of the intersection-and-joint ratio of a vehicle target history or the coefficient of variation of the speed of the history, j represents the corresponding intersection-and-joint ratio or speed, Indicates the maximum value of the intersection-over-combination coefficient of variation of all vehicle target histories or the speed coefficient of variation of the history in the current frame, A iou , A vel They represent the adaptive coefficient of the intersection-and-joint ratio and the adaptive coefficient of the speed of a certain vehicle target history, respectively. i init represents the initialization weight;

[0199] In this embodiment, the initialization weights are preferably set to

[0200] S63, the weight w obtained by adaptive adjustment i Substitute the weighted vehicle matching probability Prob ego The calculation formula is used to obtain the weighted self-vehicle matching probability Prob of the corresponding tracking sequence belonging to the vehicle itself after adaptive weighted calculation. ego , take the weighted self-vehicle matching probability Prob ego The target position of the highest vehicle target is used as the current position of the vehicle itself.

[0201] Compared with the prior art, the beneficial effects of this embodiment 2 are:

[0202] This embodiment relies on roadside perception information from various types of sensor detection sources, the vehicle's own geometric dimensions and speed information to perform vehicle-road collaborative positioning. Compared with the existing positioning method based on vehicle-to-vehicle collaboration, this embodiment can get rid of the dependence on GNSS signals. Compared with the existing vehicle-to-road collaborative positioning method based on the roadside RSU receiving signal strength, this embodiment can achieve long-distance continuous positioning of networked vehicles and obtain the relative position relationship with surrounding vehicles, which is more practical.

[0203] Furthermore, this embodiment, through the step of target position prediction, can predict the position information of the vehicle of the currently tracked target based on historical data when the roadside perception information is lost. Compared with a single GPS positioning or a single roadside perception positioning method, it can effectively increase the positioning signal output frequency and provide more real-time positioning information for the connected vehicle.

[0204] Furthermore, this embodiment, through the step of self-vehicle probability prediction, combined with an adaptive weighting method based on coefficient of variation statistics, can effectively locate the vehicle's own position information when connected vehicles and non-connected vehicles are mixed, has high positioning accuracy and positioning precision, and has practical application value.

[0205] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. A vehicle-road collaborative positioning system, characterized in that: It includes a roadside sensor unit, an edge computing unit, a roadside communication unit, a vehicle communication unit, and a vehicle computing unit that are sequentially connected for data transmission; The roadside sensor unit is used to collect information about vehicles in the road area; The edge computing unit is used to process information collected by the roadside sensor unit to form roadside perception information of vehicles in the road area; The roadside communication unit is used to send the roadside perception information processed by the edge computing unit to the vehicle-mounted communication unit; The vehicle-mounted communication unit is used to transmit the received roadside sensing information to the vehicle-mounted computing unit; The vehicle-mounted computing unit is used to locate the vehicle itself in the road area according to the roadside perception information and the vehicle's own state information; The vehicle-mounted computing unit includes a pre-processing module, a multi-target matching and tracking module, a target position prediction module, and an ego-vehicle probability prediction module which are connected in sequence; The preprocessing module is connected to the vehicle communication unit and is used to preprocess the roadside sensing information and filter out invalid non-vehicle targets and repeated vehicle targets; The multi-target matching and tracking module is used to obtain and track the position and speed changes of vehicle targets according to roadside perception information; The target position prediction module is used to predict the roadside perception information and correct the position of the vehicle target according to the position and speed changes of the vehicle target using an extended Kalman filter, and then update the roadside perception information; The vehicle probability prediction module is used to calculate the probability that the currently tracked vehicle target is the vehicle itself according to the updated roadside perception information, so as to realize the positioning of the vehicle; The roadside sensor unit includes a plurality of sensors deployed at a plurality of points in the road area, and the types of the sensors include laser radar and camera; The laser radar and camera are used to obtain point cloud and image data of the road area; Each sensor is electrically connected to an edge computing unit for data transmission, and each sensor is provided with global positioning coordinates and joint calibration parameters corresponding to the current coordinate system of the road area; The edge computing unit includes a target detection module, a perception fusion module, a coordinate conversion module, and an encoding module; The target detection module is connected to multiple sensors in the roadside sensor unit, and is used to perform two-dimensional target detection on the image and point cloud data collected by the roadside multiple sensors, and output target detection information of different sensors; the target detection information includes the geometric size, position, and detection frame of the vehicle target; The perception fusion module is connected to the target detection module, and is used to perform data matching, association, and misdetection elimination processing on the image two-dimensional target detection information and the point cloud three-dimensional target detection information, and output the processed three-dimensional target detection information as roadside perception information; The coordinate conversion module is connected to the perception fusion module and is used to convert the coordinate points in the current coordinate system of the road area into the latitude and longitude coordinate target detection information in the world coordinate system; The encoding module is connected to the coordinate conversion module and is used to encode the roadside perception information according to the standard C-V2X data communication protocol; The preprocessing module is used to receive the encoded roadside perception information, decode the roadside perception information, convert the longitude and latitude coordinates of the world coordinate system in the roadside perception information into the coordinates of the current road area coordinate system, and remove invalid data of non-vehicle targets and repeated detection targets according to the roadside perception information; The process of obtaining and tracking the position and speed changes of the vehicle target is as follows: Traverse the roadside perception information of all detected vehicle targets and all tracked vehicle targets in the current frame, and set them as detection targets and tracking targets respectively; If the current frame is the first frame to start tracking, each detected target in the current frame will be directly used as the tracking target of the next frame; If the current frame is not the first frame to start tracking, the geometric size intersection and union ratio, position distance, and speed derived from the position distance and the time difference between the previous and next frames of each detected target in the current frame and all tracked targets are calculated; The geometric dimension intersection-union ratio is calculated by aligning the geometric center points of the detection frames of the detection target and the tracking target, and calculating the ratio of the intersection volume to the union volume of the two detection frames; The position distance is calculated by: calculating the Euclidean distance between the geometric center points of the detection frame of the detection target and the tracking target; The speed is calculated by dividing the distance between the detection frame of the detection target and the tracking target by the time interval between the previous and next frames; Set the speed matching probability Prob vel And the calculation is shown as follows: Among them, v c 、v t 、v thresh Respectively represent the speed of the detected target in the current frame, the speed of the tracked target, and the set speed threshold; Set the distance matching probability Prob dis And the calculation is shown as follows: Among them, d o d thresh They represent the distance between the detected target and the tracked target in the current frame and the set distance threshold, respectively. Δt is the time interval between the current frame and the previous frame. Set the geometric size intersection ratio to the IOU matching probability Prob iou ; According to the IOU matching probability Prob iou , distance matching probability Prob dis , speed matching probability Prob vel , calculate the target matching probability Prob of the detected target and all tracked targets in the current frame m , target matching probability Prob m As shown below: Prob m =w iou Prob iou +w vel Prob vel +w dis Prob dis Among them, w iou 、w vel 、w dis Corresponding to IOU matching probability Prob iou , speed matching probability Prob vel , distance matching probability Prob dis The respective weight values; After all the detection targets in the current frame have been traversed accordingly, according to the target matching probability Prob m Update the roadside perception information of all tracked targets.

2. The vehicle-road collaborative positioning system according to claim 1, characterized in that: According to the target matching probability Prob m The process of updating the roadside perception information of all tracked targets is as follows: If there is a detection target corresponding to part or all of the tracking targets, the target matching probability Prob m If it is greater than the set matching threshold, the target matching probability Prob is judged m The highest corresponding tracking target and the detection target in the current frame are the same vehicle, and then the roadside perception information of the detection target in the current frame is used to replace the roadside perception information of the corresponding tracking target; If there is a detection target in the current frame that corresponds to all tracked targets, the target matching probability Prob m If the value is less than the set matching threshold, it is determined that the detected target cannot find the corresponding tracking target in the previous frame, and then the roadside perception information of the detected target is directly used as the roadside perception information of the corresponding tracking target in the next frame.

3. The vehicle-road collaborative positioning system according to claim 2, characterized in that: The process of using the extended Kalman filter to predict the roadside perception information and correct the position of the vehicle target is as follows: If there is a detected target in the current frame that cannot be matched with the tracked target, an extended Kalman filter is constructed and initialized using the position of the detected target in the current frame; If there is a detection target in the current frame that matches a certain tracking target, the extended Kalman filter corresponding to the tracking target is predicted using the position and speed corresponding to the detection target in the current frame, and the corrected position is output as the latest position of the detection target; If there is a tracking target that cannot match the detection target of the current frame, the roadside perception information of the tracking target and its extended Kalman filter are used, combined with the time interval between the current frame and the previous frame, to predict and update the position of the tracking target corresponding to the current frame.

4. The vehicle-road collaborative positioning system according to claim 3, characterized in that: The process of calculating the probability that the currently tracked vehicle target is the vehicle itself is as follows: Traverse the roadside perception information of all vehicle targets in the current frame after the update, and statistically calculate the geometric size, position, speed, and moved position distance of each vehicle target; According to the geometric size of the vehicle itself, the speed of the vehicle itself, and the position and distance information of the vehicle itself, the latest intersection and union ratio of the detection frame of the vehicle itself and all vehicle targets is calculated respectively. Average Intersection-Union Ratio The probability that the vehicle target belongs to the vehicle itself Probability of the vehicle moving distance Then, based on the adaptive weighted method of coefficient of variation statistics, the weighted self-vehicle matching probability Prob that the vehicle target belongs to the vehicle itself is calculated ego , according to the weighted self-car matching probability Prob ego Determine whether the vehicle target corresponding to the corresponding tracking sequence belongs to the vehicle itself; The vehicle speed probability Probability of the vehicle moving distance Weighted vehicle matching probability Prob ego The calculation formula is as follows: Among them, v t 、v e Respectively represent the speed of the vehicle target and the speed of the vehicle itself, d t d e d thresh They represent the moving distance of the vehicle target, the moving distance of the vehicle, and the moving distance threshold, respectively. i represents the weight, Represents the weighted item corresponding to type i.

5. The vehicle-road collaborative positioning system according to claim 4, characterized in that: The adaptive weighting method based on coefficient of variation statistics includes: Compare the latest intersection Average Intersection-Union Ratio Vehicle speed probability Probability of the vehicle moving distance After the weights are initialized, the intersection-over-combination ratio variation coefficient and the speed variation coefficient are set to perform adaptive adjustment on different weights. The calculation of adaptive adjustment is shown in the following formula: Among them, CV j represents the coefficient of variation of the intersection-and-joint ratio of a vehicle target history or the coefficient of variation of the speed of the history, j represents the corresponding intersection-and-joint ratio or speed, Indicates the maximum value of the historical intersection-over-combination coefficient of variation or the historical speed coefficient of variation among all vehicle targets in the current frame, A iou , A vel They represent the historical IoU adaptive coefficient of a vehicle target and the historical speed adaptive coefficient of init respectively, i init Represents the initialization weights.

6. A vehicle-road collaborative positioning method, characterized in that: The steps include: Collect information about vehicles in the road area; Processing the collected information to form roadside perception information of vehicles in the road area; Pre-process the roadside perception information of each vehicle to filter out invalid non-vehicle targets and repeated vehicle targets; Based on the roadside perception information of multiple vehicles, the position and speed changes of vehicle targets are obtained and tracked; According to the position and speed changes of the vehicle target, the extended Kalman filter is used to predict and correct the position of the roadside perception information set, and then update the roadside perception information; The probability that the currently tracked vehicle target is the vehicle itself is calculated based on the updated roadside perception information, thereby realizing the positioning of the vehicle; The process of obtaining and tracking the position and speed changes of vehicle targets is as follows: Traverse the roadside perception information of all detected vehicle targets and all tracked vehicle targets in the current frame, and set them as detection targets and tracking targets respectively; If the current frame is the first frame to start tracking, each detected target in the current frame will be directly used as the tracking target of the next frame; If the current frame is not the first frame to start tracking, the geometric size intersection and union ratio, position distance, and speed derived from the position distance and the time difference between the previous and next frames of each detected target in the current frame and all tracked targets are calculated; The geometric dimension intersection-union ratio is calculated by aligning the geometric center points of the detection frames of the detection target and the tracking target, and calculating the ratio of the intersection volume to the union volume of the two detection frames; The position distance is calculated by: calculating the Euclidean distance between the geometric center points of the detection frame of the detection target and the tracking target; The speed is calculated by dividing the distance between the detection frame of the detection target and the tracking target by the time interval between the previous and next frames; Set the speed matching probability Prob vel And the calculation is shown as follows: Among them, v c 、v t 、v thresh Respectively represent the speed of the detected target in the current frame, the speed of the tracked target, and the set speed threshold; Set the distance matching probability Prob dis And the calculation is shown as follows: Among them, d o d thresh They represent the distance between the detected target and the tracked target in the current frame and the set distance threshold, respectively. Δt is the time interval between the current frame and the previous frame. Set the geometric size intersection ratio to the IOU matching probability Prob iou ; According to the IOU matching probability Prob iou , distance matching probability Prob dis , speed matching probability Prob vel , calculate the target matching probability Prob of the detected target and all tracked targets in the current frame m , target matching probability Prob m As shown below: Prob m =w iou Prob iou +w vel Prob vel +w dis Prob dis Among them, w iou 、w vel 、w dis Corresponding to IOU matching probability Prob iou , speed matching probability Prob vel , distance matching probability Prob dis The respective weight values; After all the detection targets in the current frame have been traversed accordingly, according to the target matching probability Prob m Update the roadside perception information of all tracked targets.

Images