Search

CN-121982694-A - ETC and vehicle perception target point location matching method based on charging portal multi-mode perception

CN121982694ACN 121982694 ACN121982694 ACN 121982694ACN-121982694-A

Abstract

The invention discloses an ETC and vehicle perception target point location matching method based on charging portal multi-mode perception, and relates to the technical field of vehicle perception matching. The method comprises the steps of obtaining position information and quantity information of a highway toll gate, obtaining an obtained information set, installing a perception system and an ETC system based on the obtained information set, and obtaining perception data of a vehicle based on the perception system. According to the invention, a multi-modal sensing system is constructed, a deep learning model is introduced to optimize, a point location matching algorithm is utilized to perform space-time alignment and fusion on normalized ETC data and sensing data, a deep learning model based on a convolutional neural network is introduced, characteristics such as longitude and latitude, course angle and time stamp after fusion are input, matching probability and predicted track are output, the limitation of a single vision system is greatly overcome, and vehicle positioning accuracy and license plate recognition success rate under complex expressway environments such as illumination change, shielding, long distance and pollution are remarkably improved.

Inventors

  • SHEN YANWEI

Assignees

  • 成都通广网联科技有限公司

Dates

Publication Date
20260505
Application Date
20260407

Claims (8)

  1. 1. ETC and vehicle perception target point matching method based on charging portal multi-mode perception comprises the following steps: acquiring position information and quantity information of a highway toll gate, acquiring an acquired information set, installing a perception system and an ETC system based on the acquired information set, acquiring perception data of a vehicle based on the perception system, acquiring a perception data set, and acquiring marking information of the vehicle based on the ETC system to acquire a vehicle information item; Characterized by further comprising: based on the combination result of the vehicle information item and the perception data set, realizing clock synchronization through local area network connection to obtain an information synchronization set; matching according to the vehicle information item and the sensing data set by utilizing a point location matching algorithm, and synchronizing ETC license plate parameters to a sensing system when the matching is successful, wherein the ETC license plate parameters are used for license plate supplementation and redundancy judgment; Setting an installation gap, installing a plurality of perception subsystems based on the installation gap after a charging portal, wherein each subsystem comprises millimeter wave radars and an edge calculation unit, performing target tracking through a coverage overlapping area between the millimeter wave radars, generating unique participant IDs in a matching way, inheriting license plate information, and obtaining matching information items; And introducing a deep learning model for optimizing point location matching and calibration algorithm, wherein the deep learning model is trained based on a convolutional neural network, inputs the parameters including the first longitude and latitude parameter, the second longitude and latitude parameter, the course angle information and the time stamp parameter, and outputs the parameters as matching probability and a predicted track so as to improve license plate recognition accuracy under the condition of poor sight.
  2. 2. The ETC and vehicle perception target point matching method based on charging portal multi-mode perception according to claim 1, wherein the acquisition method for acquiring the information set comprises the following steps: Acquiring data based on an ETC system database of a highway management department and an intelligent transportation platform, wherein the data comprises a portal ID, longitude and latitude coordinates, an installation date and a portal state, so as to obtain a portal information set; performing portal screening based on the portal information set to obtain corresponding portal information in an enabled state, thereby obtaining a screening information set; Setting a screening threshold, wherein the screening threshold is a fixed time value, screening and updating the portal information set and the screening information set based on the screening threshold, further obtaining an updated screening information set, and taking the updated screening information set as an acquired information set.
  3. 3. The ETC and vehicle perception target point matching method based on charging portal multi-mode perception according to claim 1, wherein the perception system comprises a vehicle-road cooperative camera, a millimeter wave radar and an edge computing unit, and the acquisition method of the perception data set comprises the following steps: Capturing an appearance image of a vehicle based on a vehicle road cooperative camera, wherein the appearance image comprises color information and vehicle type information, and obtaining an image information item; Measuring speed parameters and running parameters of the vehicle based on the millimeter wave radar, wherein the running parameters comprise positions and course angles, and obtaining running information items; acquiring longitude and latitude information of a millimeter wave radar corresponding to a running information item, obtaining a parameter information item, and acquiring longitude and latitude information of a target vehicle based on the parameter information item to obtain a first parameter item; The image information item, the driving information item and the first parameter item are combined to obtain a perception data set.
  4. 4. The ETC and vehicle perception target point matching method based on multi-mode perception of a charging portal according to claim 3, wherein the ETC system comprises a road side unit, a high-precision positioning module and a time synchronization module, and the vehicle information item acquisition method comprises the following steps: transmitting a license plate plaintext of a target vehicle based on a vehicle-mounted electronic tag to obtain a transmission information item, performing real-time desensitization on a road side unit end based on the transmission information item, and mapping the transmission information item to a unique vehicle ID to obtain a vehicle marking item; The road side unit obtains a vehicle time item based on the accurate time of the vehicle-mounted electronic tag signal received by the time synchronization module, and the road side unit fuses the high-precision positioning module to perform auxiliary positioning to obtain a vehicle position and obtain a vehicle position item; the vehicle mark item, the vehicle time item and the vehicle position item are combined to obtain the vehicle information item.
  5. 5. The ETC and vehicle perception target point matching method based on charging portal multi-mode perception according to claim 4, wherein the perception system and the ETC system further comprise a data encryption and anonymization mechanism, and the method for acquiring the perception data set and the vehicle information item comprises the following steps: Encrypting the perception data set and the vehicle information item by adopting an end-to-end encryption technology and a setting algorithm, establishing a safe channel through a TLS protocol when communicating between a road side unit and an edge computing unit, dynamically generating an encryption key by a central server, setting a switching threshold, and rotating the encryption key based on the switching threshold; And integrating an anonymization module in the edge calculation unit, replacing the vehicle marking item with a hash value, only reserving a necessary identifier, wherein the necessary identifier comprises a vehicle type and a speed, the vehicle marking item is mapped to a unique ID, and the original data is deleted immediately after being processed, so that encryption optimization is carried out on the perception data set and the vehicle information item.
  6. 6. The ETC and vehicle perception target point matching method based on charging portal multi-mode perception according to claim 1, wherein the point matching algorithm comprises the following steps: Data alignment and standardization are carried out based on the perception data set and the vehicle information item, wherein the data alignment and standardization comprise time alignment, coordinate conversion and data cleaning, the standardized vehicle information item and the perception data set are output, and a first output information set is obtained; performing similarity calculation based on the first output information set, wherein the similarity calculation comprises spatial distance, time difference and angle difference, calculating a matching score through a similarity function, and outputting a similarity score and a related index of each target pair to obtain a second output information set; and setting a threshold value based on the similarity score, and carrying out matching judgment on the second output information set.
  7. 7. The ETC and vehicle perception target point matching method based on charging portal multi-mode perception according to claim 1, wherein the matching information item obtaining method comprises the following steps: Acquiring an effective radius of the millimeter wave radar, performing arrangement of a perception subsystem based on the effective radius to obtain at least two arrangement information items, and outputting point cloud data of a target vehicle based on the arrangement information items, wherein the point cloud data comprises position, speed and course angle; Based on the arrangement information item, performing target association by using the overlapping areas of adjacent millimeter wave radars, and generating a global unique ID through a space-time matching algorithm; and inputting license plate information synchronized at the portal, associating the portal license plate with a target through space-time matching in a first subsystem behind the portal, and binding the license plate to a unique participant ID (identity) of the target when the target carries the license plate into a coverage range of the subsystem.
  8. 8. The ETC and vehicle perception target point matching method based on charging portal multi-mode perception according to claim 1, wherein the deep learning model comprises: And defining the architecture, input characteristics and output format of the deep learning model, ensuring the matching of the adaptation points and the requirement of a calibration algorithm, simultaneously acquiring past data based on a highway portal system, respectively preprocessing and training, integrating the trained deep learning model into the existing system, and optimizing the matching and calibration process in real time for improving the license plate recognition accuracy.

Description

ETC and vehicle perception target point location matching method based on charging portal multi-mode perception Technical Field The invention relates to the technical field of vehicle perception matching, in particular to a charging portal multi-mode perception-based ETC and vehicle perception target point position matching method. Background The method is characterized in that the vehicle perception target point position matching is carried out in the expressway, and basically, the targets perceived by the sensors, such as radar and visually detected vehicles, are in one-to-one correspondence with corresponding positions or tracks in map, planning, road sides and track models, such as lanes, virtual guide lines, candidate lane center lines, time sequence track nodes and lane-level point positions, and because many control and management logics of the expressway depend on accurate lane attribution and relative positions on the lanes, such as lane-level traffic state estimation comprises flow, density, speed and the like, lane change risk estimation, safe distance estimation, event detection comprises congestion, pre-accident, reverse driving and the like, if the matching is not carried out, the system possibly takes wrong lanes or wrong mileage positions, and the decision is biased, so that the real-time perception and the point position matching of the vehicles are required. The method, the device and the electronic equipment for fusing the paths of the vehicles, which are disclosed in the patent publication No. CN115148023A, can map the vehicle perception data monitored by the perception equipment into a high-precision road network so as to determine the driving path characteristics of the driving paths of the vehicles, and on the basis, the driving path characteristics of each vehicle monitored by the adjacent point perception equipment on the same line are matched and fused according to the driving path characteristics of each vehicle monitored by the adjacent point perception equipment, so that the fused driving paths of the vehicles are obtained. According to the scheme, the vehicle running paths monitored by different sensing devices on the same line are uniformly mapped to the high-precision road network of the line, so that the path characteristics of different vehicle running paths are obtained based on the high-precision road network, and the uniformity of the vehicle running path characteristics is realized. In the using process of the technical scheme, the conventional sensing system, particularly the system relying on a single camera, is extremely easy to be disturbed by environment, the image quality is reduced due to insufficient light at night, so that license plate characters are blurred, even if the infrared light supplementing technology is adopted, the problems of uneven illumination, light reflection and the like are difficult to completely eliminate, the recognition accuracy is further influenced, the visibility is reduced in foggy days or rainy days, tiny particles in the atmosphere can scatter light, the image is blurred, the contrast is reduced, license plate images are difficult to clearly appear, in addition, the image quality is further reduced due to the fact that rainwater or fog is attached to the surface of the camera, and the recognition difficulty is further increased. Disclosure of Invention The invention aims to provide a charging portal multi-mode sensing-based ETC and vehicle sensing target point matching method so as to solve the problems in the background art. In order to achieve the purpose, the invention provides the following technical scheme that the ETC and vehicle perception target point matching method based on the multi-mode perception of the charging portal comprises the following steps: acquiring position information and quantity information of a highway toll gate, acquiring an acquired information set, installing a perception system and an ETC system based on the acquired information set, acquiring perception data of a vehicle based on the perception system, acquiring a perception data set, and acquiring marking information of the vehicle based on the ETC system to acquire a vehicle information item; based on the combination result of the vehicle information item and the perception data set, realizing clock synchronization through local area network connection to obtain an information synchronization set; matching according to the vehicle information item and the sensing data set by utilizing a point location matching algorithm, and synchronizing ETC license plate parameters to a sensing system when the matching is successful, wherein the ETC license plate parameters are used for license plate supplementation and redundancy judgment; Setting an installation gap, installing a plurality of perception subsystems based on the installation gap after a charging portal, wherein each subsystem comprises millimeter wave radars and an edge calculation unit, performi