Vehicle positioning method, device and related equipment

Abstract

The application provides a vehicle positioning method and device and related equipment, and relates to the technical field of artificial intelligence. After obtaining communication data in the vehicle driving process and image data in the vehicle driving process, the vehicle is positioned based on a machine learning model using the communication data and the image data, so as to obtain a first positioning result based on a first base station, and a third positioning result of the vehicle is generated by fusion calculation according to a plurality of second positioning results corresponding to a plurality of second base stations. In the vehicle positioning process, the communication conditions of the plurality of base stations and the vehicle and the driving conditions of the vehicle are fully considered, and the vehicle positioning accuracy is improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, specifically to a vehicle positioning method, device, and related equipment. Background Technology

[0002] In the field of intelligent transportation, with the rapid development of autonomous driving and vehicle-road cooperative technologies, the demand for high-precision, low-latency vehicle positioning in high-speed operating scenarios is becoming increasingly urgent. Currently, vehicle positioning is typically achieved through 5G base stations or edge computing devices. However, this method suffers from poor positioning accuracy in high-speed scenarios and complex environments. Summary of the Invention

[0003] This application provides a vehicle positioning method, device, and related equipment, which solves the problem of poor vehicle positioning accuracy in the prior art.

[0004] To solve the above problems, this application is implemented as follows: Firstly, this application provides a vehicle positioning method, the method comprising: The vehicle's communication data and image data collected by the vehicle are acquired. The communication data includes communication data between the vehicle and a first base station during the vehicle's operation. The first base station is the base station closest to the vehicle within the vehicle's communication range. The image data includes driving images captured during the vehicle's operation. The communication data and the image data are input into a machine learning model for localization to obtain the first localization result of the vehicle. The machine learning model is used to locate the vehicle. Multiple second positioning results corresponding to multiple second base stations are obtained. The multiple second base stations are other base stations other than the first base station within the communication range of the vehicle. The multiple second positioning results correspond one-to-one with the multiple second base stations. The first positioning result and the plurality of second positioning results are fused and calculated to obtain a third positioning result.

[0005] Optionally, acquiring the vehicle's communication data and the image data collected by the vehicle includes: The system acquires 5G and ultra-wideband signals from the fifth-generation mobile communication technology used by the vehicle to communicate with the first base station, and acquires driving images captured by cameras deployed along the vehicle's path. The driving images are those captured by the cameras when the vehicle enters the camera's field of view. The 5G signal is analyzed to obtain the signal strength attenuation slope, and the ultra-wideband signal is analyzed to obtain the distance change rate; Feature extraction is performed on the driving image to obtain multiple visual features, including lane lines, signs, and road environment. The communication data includes the signal strength attenuation slope and the distance change rate, and the image data includes the multiple visual features.

[0006] Optionally, the step of inputting the communication data and the image data into a machine learning model for localization to obtain the first localization result of the vehicle includes: Based on the multiple visual features, a detection box corresponding to the vehicle pattern in the driving image is determined, and the vehicle pattern is located inside the detection box. The coordinates of the center pixel of the detection box are converted into physical coordinates in the vehicle coordinate system of the vehicle, the physical coordinates including the X coordinate and the Y coordinate; The offset of the vehicle from the center line of the lane is calculated based on the X coordinate and the Y coordinate to obtain the lane offset. The communication data, the image data, and the lane line offset are input into the machine learning model for localization to obtain the first localization result.

[0007] Optionally, the step of inputting the communication data, the image data, and the lane offset into the machine learning model to obtain the first positioning result includes: The communication data, the image data, and the lane offset are input into the machine learning model for localization, and a third localization result is output. The environmental information obtained during vehicle operation includes at least one of the following: environmental parameters, light intensity, and scene type. The third positioning result is corrected based on the environmental information, the communication data, and the image data to obtain the first positioning result.

[0008] Optionally, the step of correcting the third positioning result based on the environmental information, the communication data, and the image data to obtain the first positioning result includes: Based on the variance values ​​of multiple historical 5G signals within the target time period, a first weight value corresponding to the 5G signal is determined, wherein the historical 5G signal is the historical communication data between the vehicle and the first base station; Based on the variance values ​​of multiple historical ultra-wideband signals within the target time period, a second weight value corresponding to the ultra-wideband signal is determined, wherein the historical ultra-wideband signal is the historical communication data between the vehicle and the first base station; Based on the detection confidence level corresponding to the image data, determine the third weight value corresponding to the image data; Determine the fourth weight value corresponding to the environmental information; Based on the first weight value, the second weight value, the third weight value, and the fourth weight value, the environmental information, the 5G signal, the ultra-bandwidth information, and the image data are weighted and calculated to obtain the correction value; The third positioning result is corrected based on the correction value to obtain the first positioning result.

[0009] Optionally, obtaining multiple second positioning results corresponding to multiple second base stations includes: Within the communication range of the vehicle, identify the plurality of second base stations in addition to the first base station; The vehicle information corresponding to the vehicle is sent to the plurality of second base stations, and the vehicle information includes at least one of the following: vehicle number, vehicle speed and vehicle predicted trajectory; The system acquires multiple second positioning data generated by the multiple second base stations based on the vehicle information to locate the vehicle.

[0010] Optionally, before inputting the communication data and the image data into the machine learning model for localization to obtain the first localization result of the vehicle, the method further includes: Obtain a training dataset, which includes multiple training samples. Each training sample includes a set of sample communication data, sample image data, and sample labels for the target vehicle during the same driving process. The sample communication data includes communication data between the target vehicle and the target base station. The sample image data includes driving images taken during the driving process of the target vehicle. The sample labels include the positioning results of the target vehicle. The initial machine learning model is trained based on the training dataset to obtain the machine learning model.

[0011] Secondly, this application provides a vehicle positioning device, the device comprising: The first acquisition module is used to acquire the vehicle's communication data and the image data collected by the vehicle. The communication data includes the communication data between the vehicle and the first base station during the vehicle's driving process. The first base station is the base station closest to the vehicle within the vehicle's communication range. The image data includes driving images captured during the vehicle's driving process. The positioning module is used to input the communication data and the image data into a machine learning model for positioning, and obtain the first positioning result of the vehicle. The machine learning model is used to locate the vehicle. The second acquisition module is used to acquire multiple second positioning results corresponding to multiple second base stations, wherein the multiple second base stations are other base stations other than the first base station within the communication range of the vehicle, and the multiple second positioning results correspond one-to-one with the multiple second base stations; The calculation module is used to perform fusion calculation on the first positioning result and the multiple second positioning results to obtain a third positioning result.

[0012] Thirdly, this application also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the method described in the first aspect above.

[0013] Fourthly, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method described in the first aspect above.

[0014] Fifthly, this application also provides a computer program product, including computer instructions that, when executed by a processor, implement the steps of the method described in the first aspect above.

[0015] This application provides a vehicle positioning method, apparatus, and related equipment, relating to the field of artificial intelligence technology. The method includes: acquiring communication data of a vehicle and image data collected by the vehicle, wherein the communication data includes communication data between the vehicle and a first base station during vehicle operation, the first base station being the base station closest to the vehicle within the vehicle's communication range, and the image data including driving images captured during the vehicle's operation; inputting the communication data and the image data into a machine learning model for positioning to obtain a first positioning result for the vehicle, the machine learning model being used to position the vehicle; acquiring multiple second positioning results corresponding to multiple second base stations, wherein the multiple second base stations are other base stations within the vehicle's communication range besides the first base station, and the multiple second positioning results correspond one-to-one with the multiple second base stations; and performing a fusion calculation on the first positioning result and the multiple second positioning results to obtain a third positioning result. The technical solution of this application, after acquiring communication data and image data during vehicle operation, uses a machine learning model to locate the vehicle using the communication data and image data, thereby obtaining a first positioning result based on a first base station. Then, based on multiple second positioning results corresponding to multiple second base stations, a third positioning result of the vehicle is generated by fusion calculation. Thus, in the vehicle positioning process, the communication between multiple base stations and the vehicle and the vehicle's driving situation are fully considered, thereby improving the accuracy of vehicle positioning. Attached Figure Description

[0016] To more clearly illustrate the technical solution of this application, the drawings used in the description of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart illustrating a vehicle positioning method provided in an embodiment of this application; Figure 2 This is a schematic diagram of the positioning process provided in an embodiment of this application; Figure 3 This is a schematic diagram of the model structure provided in the embodiments of this application; Figure 4 This is a schematic diagram of the structure of a vehicle positioning device provided in an embodiment of this application; Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] The terms "first," "second," etc., used in the embodiments of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products, or devices. Additionally, the use of "and / or" in this application indicates at least one of the connected objects, such as A and / or B and / or C, representing seven possibilities: including A alone, B alone, C alone, and the presence of both A and B, both B and C, both A and C, and the presence of A, B, and C.

[0020] See Figure 1 , Figure 1 This is a flowchart illustrating the vehicle positioning method provided in an embodiment of this application. Figure 1 As shown, the vehicle positioning method may include the following steps: Step 101: Obtain the vehicle's communication data and the image data collected by the vehicle. The communication data includes the communication data between the vehicle and the first base station during the vehicle's operation. The first base station is the base station closest to the vehicle within the vehicle's communication range. The image data includes driving images captured during the vehicle's operation.

[0021] In this embodiment, the vehicle's communication data includes communication data between the vehicle and the first base station during the vehicle's operation, such as, for example, 5G or Ultra Wide Band (UWB) signals sent by the vehicle, and data interaction with adjacent base stations and mobile reference vehicles to achieve information sharing and collaboration.

[0022] The image data collected by the vehicle includes images of the vehicle in motion. For example, images of the vehicle's surrounding environment can be collected by cameras deployed at the base station. The images are preprocessed to extract visual features such as lane lines, signs, and the surrounding environment, and finally the relative position offset is calculated.

[0023] Step 102: Input the communication data and the image data into the machine learning model for positioning to obtain the first positioning result of the vehicle. The machine learning model is used to locate the vehicle.

[0024] In this embodiment, a machine learning model is used to locate the vehicle based on communication data and image data. For example, the machine learning model can be a PyTorch CNN-LSTM hybrid model. The CNN-LSTM hybrid model is a deep learning architecture that combines a Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM) network. It is typically used to process multidimensional data with temporal characteristics, such as videos and time-series image sequences, aiming to leverage the spatial feature extraction capabilities of CNNs and the temporal series modeling capabilities of LSTMs. The structure of the CNN-LSTM model includes a convolutional neural network and a long short-term memory network. The CNN is used to extract local spatial features. When processing image sequences or video frames, each frame is first input into the CNN to capture the spatial information within the frame. Multiple convolutional layers, pooling layers, etc., can be used to progressively extract image features, thereby obtaining a set of multidimensional feature vectors. The LSTM inputs the sequence of feature vectors extracted by the CNN into the LSTM. LSTM is a special type of recurrent neural network (RNN) suitable for processing and predicting temporal data, capable of capturing long-term dependencies. In LSTM, the temporal dimension of the feature sequence is used to model temporal correlation, enabling the model to understand the dynamic changes of these features over time.

[0025] Step 103: Obtain multiple second positioning results corresponding to multiple second base stations. The multiple second base stations are other base stations other than the first base station within the communication range of the vehicle. The multiple second positioning results correspond one-to-one with the multiple second base stations.

[0026] In this embodiment, multiple second base stations are other base stations within the vehicle's communication range besides the first base station. Multiple second positioning results are generated through these multiple second base stations. Specifically, adjacent base stations and a mobile reference vehicle (such as a bus) can form a cooperative cluster to share vehicle motion status and positioning data. Data is transmitted between the multiple second base stations via a 5G NR interface (low-latency channel) with a transmission period of 100ms. Simultaneously, the mobile reference vehicle broadcasts its own position via UWB to help correct positioning errors.

[0027] Step 104: Perform a fusion calculation on the first positioning result and the multiple second positioning results to obtain a third positioning result.

[0028] In this embodiment, a modified third positioning result is obtained by comprehensively considering multiple first positioning results and multiple second positioning results.

[0029] The technical solution of this application, after acquiring communication data and image data during vehicle operation, uses a machine learning model to locate the vehicle using the communication data and image data, thereby obtaining a first positioning result based on a first base station. Then, based on multiple second positioning results corresponding to multiple second base stations, a third positioning result of the vehicle is generated by fusion calculation. Thus, in the vehicle positioning process, the communication between multiple base stations and the vehicle and the vehicle's driving situation are fully considered, thereby improving the accuracy of vehicle positioning.

[0030] In some feasible implementations, optionally, acquiring the vehicle's communication data and the image data collected by the vehicle includes: The system acquires 5G and ultra-wideband signals from the fifth-generation mobile communication technology used by the vehicle to communicate with the first base station, and acquires driving images captured by cameras deployed along the vehicle's path. The driving images are those captured by the cameras when the vehicle enters the camera's field of view. The 5G signal is analyzed to obtain the signal strength attenuation slope, and the ultra-wideband signal is analyzed to obtain the distance change rate; Feature extraction is performed on the driving image to obtain multiple visual features, including lane lines, signs, and road environment. The communication data includes the signal strength attenuation slope and the distance change rate, and the image data includes the multiple visual features.

[0031] In this embodiment, as Figure 2 As shown, Figure 2 This is a schematic diagram of the positioning process in this embodiment. When the vehicle enters the coverage area of ​​the base station, the system automatically triggers the signal acquisition and preprocessing process. The 5G signal can be connected to the baseband chip through a high-speed hardware interface (such as PCIe). Adaptive Kalman filtering is used to analyze the 5G signal and UWB signal in real time, extracting the signal strength attenuation slope (reflecting the vehicle's movement trend) from the 5G signal and calculating the distance change rate from the UWB signal. Specifically, the multimode communication module receives the vehicle signal, analyzes the 5G signal and UWB signal, and performs adaptive Kalman filtering on the received signal strength RSSI and time difference of arrival TDOA extracted from the 5G signal and the ToF timestamp extracted from the UWB signal to eliminate multipath noise.

[0032] Images of the vehicle's surrounding environment are captured by cameras deployed at the base station. These images are preprocessed to extract visual features such as lane lines, signs, and the surrounding environment. Finally, the relative positional offset is calculated. Specifically, first, distortion correction (based on the camera intrinsic matrix) and ROI cropping (focusing on the lane area) are performed. Then, visual feature matching is performed based on a lightweight YOLOv5 model, converting visual elements such as lane lines, traffic signs, and adjacent vehicles into the pixel coordinates (x, y) of the center point of a detection box containing confidence and category labels. Finally, using the center point of the lane line detection box as a reference, the lateral offset of the vehicle's center point (preset camera installation position) is calculated. The vision module captures vehicle images, and with the help of the lightweight YOLOv5 model, the system can quickly extract key visual features such as lane lines and traffic signs, and calculate the vehicle's positional offset relative to the lane lines.

[0033] In this embodiment, the acquired communication data includes the signal strength attenuation slope and the distance change rate, and the acquired image data includes the multiple visual features.

[0034] Optionally, the step of inputting the communication data and the image data into a machine learning model for localization to obtain the first localization result of the vehicle includes: Based on the multiple visual features, a detection box corresponding to the vehicle pattern in the driving image is determined, and the vehicle pattern is located inside the detection box. The coordinates of the center pixel of the detection box are converted into physical coordinates in the vehicle coordinate system of the vehicle, the physical coordinates including the X coordinate and the Y coordinate; The offset of the vehicle from the center line of the lane is calculated based on the X coordinate and the Y coordinate to obtain the lane offset. The communication data, the image data, and the lane line offset are input into the machine learning model for localization to obtain the first localization result.

[0035] In this embodiment, a detection box corresponding to the vehicle image is determined in the driving image. Specifically, the detection box needs to completely include the vehicle pattern. In addition, the detection box may also include other identifying visual features such as lane lines and signs.

[0036] Specifically, the pixel coordinates of the detection box center point are converted to physical coordinates (unit: meters) in the vehicle coordinate system, using the following formula: X = (x_pixel - cx) * kx Y = (y_pixel - cy) * ky Where cx and cy are the center points of the image, kx = 0.05m / pixel and ky = 0.1m / pixel, based on the camera calibration parameters. x_pixel represents the horizontal coordinate pixel value of the center point of the lane detection box (range 0~640, based on the 640×640 input resolution of YOLOv5s), and y_pixel represents the vertical coordinate pixel value of the center point of the lane detection box.

[0037] Using the center point of the lane line detection frame as a reference, calculate the lateral offset Δd of the vehicle center point (preset camera installation position): Δd = X_lane - X_ego Where X_lane represents the lateral physical coordinates of the lane line in the vehicle coordinate system (unit: meters), which is obtained by converting pixel coordinates; X_ego represents the preset lateral coordinates of the vehicle center point in the vehicle coordinate system (default X_ego=0, i.e., the camera installation position).

[0038] Optionally, the step of inputting the communication data, the image data, and the lane offset into the machine learning model to obtain the first positioning result includes: The communication data, the image data, and the lane offset are input into the machine learning model for localization, and a third localization result is output. The environmental information obtained during vehicle operation includes at least one of the following: environmental parameters, light intensity, and scene type. The third positioning result is corrected based on the environmental information, the communication data, and the image data to obtain the first positioning result.

[0039] In this embodiment, when using a machine learning model for localization, the communication data, the image data, and the lane line offset are first input into the machine learning model for localization, and a third localization result is output.

[0040] The system also includes environmental information during vehicle operation, which includes at least one of the following: environmental parameters, light intensity, and scene type. Specifically, environmental parameters include rainfall intensity (mm / h), light intensity is lux, and scene type includes 0-open space, 1-tunnel, 2-urban canyon, etc., which are not specifically limited in this embodiment.

[0041] Table 1 shows the calculated weights for the environmental parameters (static features, updated synchronously every frame): Rainfall intensity (mm / h); Light intensity (lux); Scene type encoding (0-open, 1-tunnel, 2-city canyon).

[0042] Table 1. Mapping Rules for Environmental Parameters and Weights

[0043] Optionally, the step of correcting the third positioning result based on the environmental information, the communication data, and the image data to obtain the first positioning result includes: Based on the variance values ​​of multiple historical 5G signals within the target time period, a first weight value corresponding to the 5G signal is determined, wherein the historical 5G signal is the historical communication data between the vehicle and the first base station; Based on the variance values ​​of multiple historical ultra-wideband signals within the target time period, a second weight value corresponding to the ultra-wideband signal is determined, wherein the historical ultra-wideband signal is the historical communication data between the vehicle and the first base station; Based on the detection confidence level corresponding to the image data, determine the third weight value corresponding to the image data; Determine the fourth weight value corresponding to the environmental information; Based on the first weight value, the second weight value, the third weight value, and the fourth weight value, the environmental information, the 5G signal, the ultra-bandwidth information, and the image data are weighted and calculated to obtain the correction value; The third positioning result is corrected based on the correction value to obtain the first positioning result.

[0044] In this embodiment, after each localization model inference is completed, the system immediately triggers a multi-source data fusion process to further improve localization accuracy. The system dynamically allocates fusion weights based on the reliability of the signal sources and environmental conditions.

[0045] Based on the input data, the system calculates fusion weights, namely the first weight w_5G for 5G signals, the second weight w_UWB for UWB signals, the third weight w_vision for visual data, the fourth weight w-motion for the motion model, and the weight w_raw for the AI ​​model. These weights can be dynamically adjusted according to real-time data to ensure the optimality of the fusion result.

[0046] The calculation methods for the confidence scores and weights of each data source are as follows: Confidence score definition: Reflects the instantaneous reliability of each data source (0~1), and is calculated as follows: 5G signal confidence: S_5G = 1 - (RSSI_variance / RSSI_max_variance) RSSI_variance: The variance of RSSI values ​​over the past 10 frames; RSSI_max_variance: Preset maximum variance threshold (dynamically adjusted based on base station density, default 30dBm²). UWB signal confidence level: S_UWB = 1 - (ToF_std / ToF_max_std) ToF_std: The standard deviation of UWB distance measurements over the past 10 frames; ToF_max_std: Preset maximum standard deviation (0.5 meters) Visual matching degree S_vision: S_vision = (mean confidence level of lane detection) × (1 - illumination attenuation factor) Illumination attenuation factor = max(0, 1 - current illuminance / 10^5 lux) Motion model S_motion: S_motion = 1 / (1 + exp(-10·prediction error)) Prediction error: the Euclidean distance between the predicted position and the actual position obtained through Kalman filtering; k: Scaling factor (default k=10, Error≈0.5 when the error is 0.1 meters) Initial positioning coordinates S_raw:

[0047] Numerator: Independent location estimates of each signal source (P5G, PUWB, ...) multiplied by their confidence weights Denominator: Sum of confidence levels Weight generation: Input the confidence scores corresponding to environmental features into the LSTM model, and output normalized weights. w_i = softmax(S_5G, S_UWB, S_vision,S_raw, S_motion) Coordinates_i: Coordinate values ​​calculated independently for each data source.

[0048] -5G coordinates: Coordinates calculated based on the TDOA triangulation algorithm; -UWB coordinates: Polygonal positioning coordinates based on ToF ranging; - Visual coordinates: Vehicle coordinates calculated based on lane line offset Δd and the absolute position of the base station; - Motion model coordinates: Coordinates inferred from the trajectory calculated by integrating vehicle speed and IMU data; -X_raw: Direct inference results from the AI ​​model, providing a localization reference independent of traditional signal sources.

[0049] Compensation Term: Based on the trajectory predicted by the vehicle's kinematics model, the compensation term compensates for signal delays caused by high-speed movement. This compensation mechanism effectively reduces positioning errors caused by high-speed vehicle movement, further improving positioning accuracy.

[0050] Kalman gain calculation: K = P_pred · H^T / (H·P_pred·H^T + R) Ppred: Prediction error covariance; R: Measurement noise covariance (dynamically adjusted by signal confidence). H: Observation matrix (maps the state to the measurement space) T: Timestamp (used for trajectory prediction and data synchronization) Compensation item generation: Compensation term = K · (X_earlier_predicted - X_earlier_fused) X_earlier_predicted: The coordinates predicted for the previous moment based on a vehicle kinematics model (such as a constant speed model). The position is calculated using the vehicle's real-time motion parameters (velocity v, acceleration a, heading angle θ).

[0051] X_earlier_fused: The optimal weighted fusion result of the multi-source data at the previous moment, representing the comprehensive localization estimate based on real-time signals and model inference.

[0052] The difference between X_earlier_fused, the "current coordinate," and the predicted coordinate X_earlier_predicted reflects the impact of signal delay and motion abrupt changes during previous calculations, which is dynamically compensated by the Kalman gain K.

[0053] By introducing a compensation term, the variance of the positioning error can be significantly reduced: Assuming the fusion error variance is σ_{fused}^2 and the prediction error variance is σ_{pred}^2, then the compensated error variance is: σ_{final}^2 = (1 - K)²·σ_{fused}^2 + K²·σ_{pred}^2 Therefore, when the Kalman gain K is at its optimal value (K=σ_{fused}^2 / (σ_{fused}^2+σ_{pred}^2)), the error variance is minimized. In high-speed scenarios (where σ_{pred}^2 is larger), the compensation term reduces the error variance by 30%-50%.

[0054] Redundancy design verification (signal complementarity): 5G offers wide coverage but low accuracy, while UWB offers high accuracy but limited range. Visual recognition is interference-resistant but dependent on lighting conditions. By mapping environmental parameters and weights, visual weights are turned off in rainy weather and 5G weights are turned off in tunnel scenarios. Positioning is maintained through UWB and motion models.

[0055] This embodiment achieves improved dynamic environmental adaptability and positioning accuracy. In high-speed mobile scenarios, it can optimize the positioning algorithm in real time to adapt to dynamic interference such as signal fluctuations and weather changes. For existing single-data source errors (1.5 meters for 5G; 0.1 meters for UWB; 0.3 meters for visual errors), this proposal can effectively reduce positioning errors and achieve sub-meter positioning accuracy through a dynamic weight adjustment algorithm based on multi-source data fusion.

[0056] Optionally, obtaining multiple second positioning results corresponding to multiple second base stations includes: Within the communication range of the vehicle, identify the plurality of second base stations in addition to the first base station; The vehicle information corresponding to the vehicle is sent to the plurality of second base stations, and the vehicle information includes at least one of the following: vehicle number, vehicle speed and vehicle predicted trajectory; The system acquires multiple second positioning data generated by the multiple second base stations based on the vehicle information to locate the vehicle.

[0057] In this embodiment, the vehicle ID, speed, and predicted trajectory are sent to multiple adjacent second base stations via the 5G NR interface.

[0058] Predicted trajectory acquisition method: Motion parameters are extracted from vehicle signals (5G / UWB) received by the multi-mode communication module.

[0059] Speed ​​v: v = (Δf · c) / (f_c · cosθ) Δf: Frequency offset of the received signal; f_c: 5G carrier frequency (e.g., 3.5GHz); θ: Directional angle of the line connecting the vehicle and the base station (estimated by beamforming of the base station array antenna). c: speed of light (used for UWB ranging calculation, c=3×10^8 m / s).

[0060] Predicted location: (t+Δt) = current position + v·Δt·[cosθ, sinθ] Δt: Prediction period (default 100ms), consistent with the synchronization period between base stations.

[0061] The system receives data from neighboring base stations, corrects the positioning results of this base station, and calculates the vehicle's theoretical coordinates (X_theoretical) in the local base station coordinate system based on the overlap between the predicted trajectory and the coverage area of ​​this base station. If the error between the theoretical coordinates and the actual detected signal (such as UWB ranging) exceeds a threshold (e.g., 1 meter), collaborative correction is triggered. X_corrected = α·X_fused + (1-α)·X_neighbor X_corrected: is the globally optimal coordinate after fusing multi-source data from this base station, neighboring base station positioning, and mobile reference nodes. It does not directly include 5G / UWB / visual / motion model coordinates, but is a secondary optimization of the multi-source fusion result (X_fused), which incorporates external collaborative data.

[0062] X_fused: Theoretical coordinates with errors. It is the weighted fusion result of the initial positioning coordinates X_raw and multi-source data, representing the theoretical optimal estimate.

[0063] X_neighbor: Calculated coordinates of neighboring base stations.

[0064] α: Weight, dynamically adjusted based on the distance between base stations (the closer the distance, the larger α; default α=0.7). After calculating X_fused, the positioning error is corrected by referencing the high-precision positions of mobile reference vehicles such as buses and logistics vehicles, which are broadcast via UWB. The specific operation is as follows: Base station reception and comparison: The base station measures the UWB distance d_measured between the bus and the UWB distance. Calculate the theoretical distance d_theoretical = |X_base_station - X_bus|; Calculate the distance error Δd = d_measured - d_theoretical.

[0065] Error compensation is performed proportionally, and the weighted average positioning result (X_fused) is calibrated a second time to eliminate base station deployment errors. X_final = X_fused + (Δd / d_theoretical) · (X_fused - X_bus) X_final: Corrected coordinates X_fused: Error coordinates Δd: Distance error d_theoretical: theoretical distance X_bus: Reference vehicle high-precision coordinates X_base_station: High-precision coordinates of the reference base station After the multi-source data fusion calculation is completed and X_corrected is achieved, the system immediately triggers the positioning result output and early warning process. Through the 5G URLLC (Ultra-Reliable Low-Latency Communication) channel, the base station sends high-precision positioning results to the vehicle, with UWB ranging accuracy of ±10cm, visual-assisted correction of ±20cm, and 5G signal optimization accuracy of ±1 meter. The actual error is compressed to the sub-meter level through the fusion algorithm.

[0066] The system monitors the vehicle's driving status in real time. If it detects the vehicle deviating from its lane (the visual analysis module detects that the distance between the vehicle and the lane line exceeds a set threshold) or if there is an obstacle equipped with a UWB tag ahead (the UWB signal detects that the distance to the obstacle ahead is less than a safe distance, such as 2 meters), it immediately triggers a millisecond-level warning. The warning information is quickly sent to the vehicle via the 5G URLLC channel, reminding the driver or the autonomous driving system to take emergency measures to avoid accidents. This warning mechanism can react in a very short time, effectively improving vehicle driving safety.

[0067] Through a distributed collaborative network, when high-speed vehicle movement causes base station switching, the continuity of positioning coverage is improved from 70% to 99% compared to related technologies, thus avoiding positioning loss.

[0068] Optionally, before inputting the communication data and the image data into the machine learning model for localization to obtain the first localization result of the vehicle, the method further includes: Obtain a training dataset, which includes multiple training samples. Each training sample includes a set of sample communication data, sample image data, and sample labels for the target vehicle during the same driving process. The sample communication data includes communication data between the target vehicle and the target base station. The sample image data includes driving images taken during the driving process of the target vehicle. The sample labels include the positioning results of the target vehicle. The initial machine learning model is trained based on the training dataset to obtain the machine learning model.

[0069] In this embodiment, the structure of the machine learning model is as follows: Figure 3As shown, the Softmax constraint ensures that the output weights w_5G, w_UWB, w_vision, w_raw, and w-motion satisfy Σw_i=1. The machine learning model is trained, and each time a vehicle's positioning request is received, the system immediately triggers positioning model inference; every 10 minutes, the system automatically triggers local training to continuously optimize model performance. During the inference phase: the current model is loaded (the initial model is downloaded from the cloud), input signals and visual data (preprocessed 5G signals, UWB signals, and lane offset Δd), and the initial positioning coordinates of the vehicle in the base station coordinate system (e.g., the vehicle's position in the base station coordinate system) are output.

[0070] The initial machine learning model is trained using a training dataset. Each training sample includes a set of sample communication data, sample image data, and sample labels for the target vehicle during the same driving process. It should be noted that the set of sample communication data, sample image data, and sample labels during the same driving process refers to the sample communication data, sample image data of the vehicle, and sample labels for locating the vehicle during the same driving process. For example, it can be communication data and image data within 1 minute.

[0071] Specifically, data preparation: training dataset, 100,000 5G / UWB / visual data with real location labels.

[0072] Model type: PyTorch-based CNN-LSTM hybrid model.

[0073] Federated learning process: Local training is performed every 10 minutes, updating the model parameters θ_local. Exchange parameters with neighboring base stations and calculate the global parameter θ_global = Σ(θ_local_i * w_i); θ: Model parameters (weight matrix and bias vector); w_i: Weight, allocated based on the historical accuracy of each base station.

[0074] Load the θ_global model, input signals and visual data, and output preliminary positioning coordinates X_raw.

[0075] Finally, the updated model parameters are synchronized to adjacent base stations via high-speed communication links to ensure the consistency and coordination of the models across the entire network, providing support for continuous vehicle positioning in cross-base station scenarios.

[0076] Model advantages: It supports multiple input sources and simultaneously processes 5G signals (temporal features), UWB ranging (spatial features), and visual data (image features). Cross-validation reduces the positioning error from 1.2 meters (pure WiFi) in existing technologies to 0.5 meters.

[0077] The technical solution of this application, after acquiring communication data and image data during vehicle operation, uses a machine learning model to locate the vehicle using the communication data and image data, thereby obtaining a first positioning result based on a first base station. Then, based on multiple second positioning results corresponding to multiple second base stations, a third positioning result of the vehicle is generated by fusion calculation. Thus, in the vehicle positioning process, the communication between multiple base stations and the vehicle and the vehicle's driving situation are fully considered, thereby improving the accuracy of vehicle positioning.

[0078] See Figure 4 , Figure 4 This is a structural diagram of the vehicle positioning method apparatus provided in the embodiments of this application. Figure 4 As shown, the vehicle positioning method device 400 includes: The first acquisition module 410 is used to acquire the vehicle's communication data and the image data collected by the vehicle. The communication data includes the communication data between the vehicle and the first base station during the vehicle's driving process. The first base station is the base station closest to the vehicle within the vehicle's communication range. The image data includes driving images captured during the vehicle's driving process. The positioning module 420 is used to input the communication data and the image data into a machine learning model for positioning to obtain the first positioning result of the vehicle. The machine learning model is used to locate the vehicle. The second acquisition module 430 is used to acquire multiple second positioning results corresponding to multiple second base stations, wherein the multiple second base stations are other base stations other than the first base station within the communication range of the vehicle, and the multiple second positioning results correspond one-to-one with the multiple second base stations; The calculation module 440 is used to perform a fusion calculation on the first positioning result and the plurality of second positioning results to obtain a third positioning result.

[0079] Optionally, the first acquisition module 410 includes: The first acquisition submodule is used to acquire 5G signals and ultra-wideband signals of the fifth generation mobile communication technology used for communication between the vehicle and the first base station, and to acquire driving images of the vehicle captured by cameras deployed along the path in which the vehicle travels. The driving images are images captured by the cameras when the vehicle enters the shooting range of the cameras. The parsing submodule is used to parse the 5G signal to obtain the signal strength attenuation slope, and to parse the ultra-wideband signal to obtain the distance change rate; The extraction submodule is used to extract features from the driving image to obtain multiple visual features, including lane lines, signs and road environment. The communication data includes the signal strength attenuation slope and the distance change rate, and the image data includes the multiple visual features.

[0080] Optionally, the positioning module includes: The first determining submodule is used to determine the detection box corresponding to the vehicle pattern in the driving image based on the multiple visual features, wherein the vehicle pattern is located inside the detection box; The conversion submodule is used to convert the coordinates of the center point pixel of the detection box into physical coordinates in the vehicle coordinate system of the vehicle, wherein the physical coordinates include X coordinates and Y coordinates; The calculation submodule is used to calculate the offset of the vehicle from the center line of the lane line based on the X coordinate and the Y coordinate, and to obtain the lane line offset. The communication data, the image data, and the lane line offset are input into the machine learning model for localization to obtain the first localization result.

[0081] Optionally, the communication data, the image data, and the lane offset are input into the machine learning model for localization to obtain the first localization result, including: The positioning unit is used to input the communication data, the image data, and the lane line offset into the machine learning model for positioning and output a third positioning result; An acquisition unit is used to acquire environmental information during vehicle operation, the environmental information including at least one of the following: environmental parameters, light intensity, and scene type; The correction unit is used to correct the third positioning result based on the environmental information, the communication data, and the image data to obtain the first positioning result.

[0082] Optionally, the correction unit includes: The first determining subunit is used to determine the first weight value corresponding to the 5G signal based on the variance values ​​of multiple historical 5G signals within the target time period, wherein the historical 5G signal is the historical communication data between the vehicle and the first base station. The second determining subunit is used to determine the second weight value corresponding to the ultra-wideband signal based on the variance value of multiple historical ultra-wideband signals within the target time period, wherein the historical ultra-wideband signal is the historical communication data between the vehicle and the first base station; The third determining subunit is used to determine the third weight value corresponding to the image data based on the detection confidence level corresponding to the image data. The fourth determining subunit is used to determine the fourth weight value corresponding to the environmental information; The weighting subunit is used to perform weighted calculations on the environmental information, the 5G signal, the ultra-bandwidth information, and the image data based on the first weight value, the second weight value, the third weight value, and the fourth weight value to obtain a correction value; The correction subunit is used to correct the third positioning result based on the correction value to obtain the first positioning result.

[0083] Optionally, the second acquisition module 430 includes: The second determining submodule is used to determine the plurality of second base stations other than the first base station within the communication range of the vehicle; The sending submodule is used to send the vehicle information corresponding to the vehicle to the plurality of second base stations. The vehicle information includes at least one of the following: vehicle number, vehicle speed, and vehicle predicted trajectory. The second acquisition submodule is used to acquire multiple second positioning data generated by the multiple second base stations based on the vehicle information to locate the vehicle.

[0084] Optional, also includes: The sample acquisition module is used to acquire a training dataset, which includes multiple training samples. Each training sample includes a set of sample communication data, sample image data, and sample labels for the target vehicle during the same driving process. The sample communication data includes communication data between the target vehicle and the target base station. The sample image data includes driving images taken during the driving process of the target vehicle. The sample labels include the positioning results of the target vehicle. The training module is used to train the initial machine learning model based on the training dataset to obtain the machine learning model.

[0085] The technical solution of this application, after acquiring communication data and image data during vehicle operation, uses a machine learning model to locate the vehicle using the communication data and image data, thereby obtaining a first positioning result based on a first base station. Then, based on multiple second positioning results corresponding to multiple second base stations, a third positioning result of the vehicle is generated by fusion calculation. Thus, in the vehicle positioning process, the communication between multiple base stations and the vehicle and the vehicle's driving situation are fully considered, thereby improving the accuracy of vehicle positioning.

[0086] This application also provides an electronic device. Please refer to [link to relevant documentation]. Figure 5The electronic device may include a processor 501, a memory 502, and a program 5021 stored in the memory 502 and capable of running on the processor 501.

[0087] When program 5021 is executed by processor 501, it can achieve the following: Figure 1 Any step in the corresponding method embodiment: The vehicle's communication data and image data collected by the vehicle are acquired. The communication data includes communication data between the vehicle and a first base station during the vehicle's operation. The first base station is the base station closest to the vehicle within the vehicle's communication range. The image data includes driving images captured during the vehicle's operation. The communication data and the image data are input into a machine learning model for localization to obtain the first localization result of the vehicle. The machine learning model is used to locate the vehicle. Multiple second positioning results corresponding to multiple second base stations are obtained. The multiple second base stations are other base stations other than the first base station within the communication range of the vehicle. The multiple second positioning results correspond one-to-one with the multiple second base stations. The first positioning result and the plurality of second positioning results are fused and calculated to obtain a third positioning result.

[0088] Optionally, acquiring the vehicle's communication data and the image data collected by the vehicle includes: The system acquires 5G and ultra-wideband signals from the fifth-generation mobile communication technology used by the vehicle to communicate with the first base station, and acquires driving images captured by cameras deployed along the vehicle's path. The driving images are those captured by the cameras when the vehicle enters the camera's field of view. The 5G signal is analyzed to obtain the signal strength attenuation slope, and the ultra-wideband signal is analyzed to obtain the distance change rate. Feature extraction is performed on the driving image to obtain multiple visual features, including lane lines, signs, and road environment. The communication data includes the signal strength attenuation slope and the distance change rate, and the image data includes the multiple visual features.

[0089] Optionally, the step of inputting the communication data and the image data into a machine learning model for localization to obtain the first localization result of the vehicle includes: Based on the multiple visual features, a detection box corresponding to the vehicle pattern in the driving image is determined, and the vehicle pattern is located inside the detection box. The coordinates of the center pixel of the detection box are converted into physical coordinates in the vehicle coordinate system of the vehicle, the physical coordinates including the X coordinate and the Y coordinate; The offset of the vehicle from the center line of the lane is calculated based on the X coordinate and the Y coordinate to obtain the lane offset. The communication data, the image data, and the lane line offset are input into the machine learning model for localization to obtain the first localization result.

[0090] Optionally, the step of inputting the communication data, the image data, and the lane offset into the machine learning model to obtain the first positioning result includes: The communication data, the image data, and the lane offset are input into the machine learning model for localization, and a third localization result is output. The environmental information obtained during vehicle operation includes at least one of the following: environmental parameters, light intensity, and scene type. The third positioning result is corrected based on the environmental information, the communication data, and the image data to obtain the first positioning result.

[0091] Optionally, the step of correcting the third positioning result based on the environmental information, the communication data, and the image data to obtain the first positioning result includes: Based on the variance values ​​of multiple historical 5G signals within the target time period, a first weight value corresponding to the 5G signal is determined, wherein the historical 5G signal is the historical communication data between the vehicle and the first base station; Based on the variance values ​​of multiple historical ultra-wideband signals within the target time period, a second weight value corresponding to the ultra-wideband signal is determined, wherein the historical ultra-wideband signal is the historical communication data between the vehicle and the first base station; Based on the detection confidence level corresponding to the image data, determine the third weight value corresponding to the image data; Determine the fourth weight value corresponding to the environmental information; Based on the first weight value, the second weight value, the third weight value, and the fourth weight value, the environmental information, the 5G signal, the ultra-bandwidth information, and the image data are weighted and calculated to obtain the correction value; The third positioning result is corrected based on the correction value to obtain the first positioning result.

[0092] Optionally, obtaining multiple second positioning results corresponding to multiple second base stations includes: Within the communication range of the vehicle, identify the plurality of second base stations in addition to the first base station; The vehicle information corresponding to the vehicle is sent to the plurality of second base stations, and the vehicle information includes at least one of the following: vehicle number, vehicle speed and vehicle predicted trajectory; The system acquires multiple second positioning data generated by the multiple second base stations based on the vehicle information to locate the vehicle.

[0093] Optionally, before inputting the communication data and the image data into the machine learning model for localization to obtain the first localization result of the vehicle, the method further includes: Obtain a training dataset, which includes multiple training samples. Each training sample includes a set of sample communication data, sample image data, and sample labels for the target vehicle during the same driving process. The sample communication data includes communication data between the target vehicle and the target base station. The sample image data includes driving images taken during the driving process of the target vehicle. The sample labels include the positioning results of the target vehicle. The initial machine learning model is trained based on the training dataset to obtain the machine learning model.

[0094] The technical solution of this application, after acquiring communication data and image data during vehicle operation, uses a machine learning model to locate the vehicle using the communication data and image data, thereby obtaining a first positioning result based on a first base station. Then, based on multiple second positioning results corresponding to multiple second base stations, a third positioning result of the vehicle is generated by fusion calculation. Thus, in the vehicle positioning process, the communication between multiple base stations and the vehicle and the vehicle's driving situation are fully considered, thereby improving the accuracy of vehicle positioning.

[0095] This application also provides a computer-readable storage medium storing a computer program. When executed by a processor, this computer program implements the various processes of the above-described vehicle positioning method embodiments and achieves the same technical effects. To avoid repetition, it will not be described again here. The computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0096] This application also provides a computer program product, which is stored in a storage medium and executed by at least one processor to implement the various processes of the above-described vehicle positioning method embodiments, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0097] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0098] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0099] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A vehicle positioning method, characterized in that, The method includes: The process involves acquiring vehicle communication data and image data collected by the vehicle. The communication data includes communication data between the vehicle and a first base station during the vehicle's operation. The first base station is the base station closest to the vehicle within its communication range. The image data includes driving images captured during the vehicle's operation. Acquiring the vehicle's communication data and image data includes: acquiring 5G and ultra-wideband signals used by the vehicle to communicate with the first base station; acquiring driving images captured by cameras deployed along the vehicle's path, wherein the driving images are images captured by the cameras when the vehicle enters the camera's field of view; analyzing the 5G signal to obtain a signal strength attenuation slope; analyzing the ultra-wideband signal to obtain a distance change rate; and extracting features from the driving images to obtain multiple visual features, including lane lines, signs, and road environment. The communication data includes the signal strength attenuation slope and the distance change rate, and the image data includes the multiple visual features. The communication data and image data are input into a machine learning model for localization to obtain a first localization result for the vehicle. The machine learning model is used to locate the vehicle. The process of inputting the communication data and image data into the machine learning model for localization to obtain the first localization result includes: determining a detection box corresponding to a vehicle pattern in the driving image based on multiple visual features, wherein the vehicle pattern is located inside the detection box; converting the coordinates of the center pixel of the detection box into physical coordinates in the vehicle's vehicle coordinate system, wherein the physical coordinates include X and Y coordinates; calculating the offset of the vehicle from the center line of the lane line based on the X and Y coordinates to obtain the lane line offset; and inputting the communication data, image data, and lane line offset into the machine learning model for localization to obtain the first localization result. Multiple second positioning results corresponding to multiple second base stations are obtained. The multiple second base stations are other base stations other than the first base station within the communication range of the vehicle. The multiple second positioning results correspond one-to-one with the multiple second base stations. The first positioning result and the plurality of second positioning results are fused and calculated to obtain a third positioning result.

2. The method according to claim 1, characterized in that, The step of inputting the communication data, the image data, and the lane offset into the machine learning model to obtain the first positioning result includes: The communication data, the image data, and the lane offset are input into the machine learning model for localization, and a third localization result is output. The environmental information obtained during vehicle operation includes at least one of the following: environmental parameters, light intensity, and scene type. The third positioning result is corrected based on the environmental information, the communication data, and the image data to obtain the first positioning result.

3. The method according to claim 2, characterized in that, The step of correcting the third positioning result based on the environmental information, the communication data, and the image data to obtain the first positioning result includes: Based on the variance values ​​of multiple historical 5G signals within the target time period, a first weight value corresponding to the 5G signal is determined, wherein the historical 5G signal is the historical communication data between the vehicle and the first base station; Based on the variance values ​​of multiple historical ultra-wideband signals within the target time period, a second weight value corresponding to the ultra-wideband signal is determined, wherein the historical ultra-wideband signal is the historical communication data between the vehicle and the first base station; Based on the detection confidence level corresponding to the image data, determine the third weight value corresponding to the image data; Determine the fourth weight value corresponding to the environmental information; Based on the first weight value, the second weight value, the third weight value, and the fourth weight value, the environmental information, the 5G signal, the ultra-bandwidth information, and the image data are weighted and calculated to obtain the correction value; The third positioning result is corrected based on the correction value to obtain the first positioning result.

4. The method according to claim 1, characterized in that, The step of obtaining multiple second positioning results corresponding to multiple second base stations includes: Within the communication range of the vehicle, identify the plurality of second base stations in addition to the first base station; The vehicle information corresponding to the vehicle is sent to the plurality of second base stations, and the vehicle information includes at least one of the following: vehicle number, vehicle speed and vehicle predicted trajectory; The system acquires multiple second positioning data generated by the multiple second base stations based on the vehicle information to locate the vehicle.

5. The method according to claim 1, characterized in that, Before inputting the communication data and the image data into the machine learning model for localization to obtain the first localization result of the vehicle, the method further includes: Obtain a training dataset, which includes multiple training samples. Each training sample includes a set of sample communication data, sample image data, and sample labels for the target vehicle during the same driving process. The sample communication data includes communication data between the target vehicle and the target base station. The sample image data includes driving images taken during the driving process of the target vehicle. The sample labels include the positioning results of the target vehicle. The initial machine learning model is trained based on the training dataset to obtain the machine learning model.

6. A vehicle positioning device, characterized in that, The device includes: A first acquisition module is used to acquire vehicle communication data and image data collected by the vehicle. The communication data includes communication data between the vehicle and a first base station during the vehicle's travel. The first base station is the base station closest to the vehicle within the vehicle's communication range. The image data includes driving images captured during the vehicle's travel. The first acquisition module includes: a first acquisition submodule, used to acquire 5G signals and ultra-wideband signals used for communication between the vehicle and the first base station, and to acquire driving images captured by cameras deployed along the vehicle's path, wherein the driving images are images captured by the cameras when the vehicle enters the camera's field of view; a parsing submodule, used to parse the 5G signals to obtain a signal strength attenuation slope, and to parse the ultra-wideband signals to obtain a distance change rate; and an extraction submodule, used to extract features from the driving images to obtain multiple visual features, including lane lines, signs, and road environment. The communication data includes the signal strength attenuation slope and the distance change rate, and the image data includes the multiple visual features. A positioning module is used to input the communication data and the image data into a machine learning model for positioning, thereby obtaining a first positioning result for the vehicle. The machine learning model is used to locate the vehicle. The positioning module includes: a first determining submodule, used to determine a detection box corresponding to a vehicle pattern in the driving image based on multiple visual features, wherein the vehicle pattern is located inside the detection box; a conversion submodule, used to convert the coordinates of the center pixel of the detection box into physical coordinates in the vehicle's vehicle coordinate system, wherein the physical coordinates include X and Y coordinates; and a calculation submodule, used to calculate the offset of the vehicle from the center line of the lane line based on the X and Y coordinates, thereby obtaining the lane line offset; and inputting the communication data, the image data, and the lane line offset into the machine learning model for positioning to obtain the first positioning result. The second acquisition module is used to acquire multiple second positioning results corresponding to multiple second base stations, wherein the multiple second base stations are other base stations other than the first base station within the communication range of the vehicle, and the multiple second positioning results correspond one-to-one with the multiple second base stations; The calculation module is used to perform fusion calculation on the first positioning result and the multiple second positioning results to obtain a third positioning result.

7. An electronic device, characterized in that, include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the method as described in any one of claims 1 to 5.

9. A computer program product, characterized in that, Includes computer instructions that, when executed by a processor, implement the steps of the method as described in any one of claims 1 to 5.

Images