Vision-based electric bus positioning model, system and related device

Abstract

The invention discloses an electric bus positioning method, system and related device based on vision, and belongs to the technical field of electric bus positioning, and the method comprises the steps: obtaining a data set with a charging station image, marking the data set, and obtaining a marked charging station image; inputting the marked charging station image into a neural network model for training to obtain a neural network model for key point detection; detecting key points on the charging station by using a key point detection neural network model; the detected key points are matched with a known charging station model, and a matched 3D-2D corresponding relation is obtained; and solving through a PnP algorithm by utilizing the matched 3D-2D corresponding relation to obtain the pose of the camera. According to the invention, rapid and accurate positioning is realized, and the efficiency of docking the charging station is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric bus positioning, and in particular relates to a vision-based electric bus positioning method, system and related devices. Background Art

[0002] As electric buses play an increasingly important role in public transportation, precise docking with charging stations has become a technical challenge, especially for drivers with insufficient driving skills. Traditional driver assistance systems (ADAS) rely on technologies such as GPS for positioning, but GPS signals are easily interfered with in urban environments, and the positioning accuracy usually cannot meet the needs of precise docking.

[0003] To solve this problem, researchers proposed a localization method based on a monocular vision system. This method assumes that the charging station is a known object and uses a monocular camera system to detect selected feature points on the charging station for localization. This method combines geometric methods to estimate the pose and deep learning models to detect key points. Specifically, the detection of key points and the initial identification of the charging station are completed by neural network models that are able to learn and identify the feature points of the charging station from the image.

[0004] The technical solution of the prior art includes the following key points: Monocular Vision System: Uses a single high-resolution camera to capture images of the charging station, avoiding the complexity of stereo vision systems that require calibration.

[0005] Feature point positioning: The camera detects feature points on the charging station, which have known positions in the 3D model of the charging station.

[0006] Neural Network Models: We propose two novel neural network architectures for keypoint estimation. These models are able to learn the features of charging stations from images and improve the accuracy of keypoint detection.

[0007] Geometric method: Using the known feature point structure and camera intrinsic parameters, the position and posture of the electric bus relative to the charging station are calculated through geometric methods.

[0008] Assisted docking: The system not only provides positioning information, but also provides clear operational prompts to drivers to help them perform the required operations to dock with the charging station. In terms of technical effects, this method is able to achieve accurate positioning of electric buses with low-cost hardware settings and minimized passive markings at charging stations. Experiments show that the proposed MRHKN (MaxResolution Heatmap Keypoint Network) architecture outperforms existing key point detectors in the tasks considered and provides the best performance in translation and rotation estimation of the bus. In addition, the system is also able to handle occasional lack of pose estimation due to occlusion or image defects (such as direct sunlight), and provides higher-frequency positioning information to the control system by integrating with vehicle odometer estimates.

[0009] In the existing technology, although GPS and DGPS can provide positioning information, they are easily affected by signal interference and occlusion in urban environments, resulting in reduced positioning accuracy and unable to meet the high-precision requirements required for electric buses to dock at charging stations. In addition, although vision-based methods, such as visual odometers and SLAM, can provide relatively accurate positioning information, they usually require a large number of significant features and perform poorly when dealing with small targets at long distances. Summary of the invention

[0010] In order to solve the above problems, the present invention proposes a vision-based electric bus positioning method, system and related devices, which achieve fast and accurate positioning and improve the efficiency of docking charging stations.

[0011] In order to solve the above technical problems, the technical solution adopted by the present invention is: In a first aspect, the present invention provides a vision-based electric bus positioning method, comprising: A data set with images of charging stations is obtained, and the data set is annotated to obtain annotated images of charging stations; Input the labeled charging station image into the neural network model for training to obtain a neural network model for key point detection; Detect key points on the charging station using a key point detection neural network model; Match the detected key points with the known charging station model to obtain a matched 3D-2D correspondence; Using the matched 3D-2D correspondence, the camera's pose is solved using the PnP algorithm.

[0012] As a further improvement of the present invention, the annotation includes key point positions on the charging station, specifically four corner points of the head of the charging station, natural corner points or artificial marking points.

[0013] As a further improvement of the present invention, the neural network model is a RKN architecture and a MRHKN architecture.

[0014] As a further improvement of the present invention, the RKN architecture consists of multiple convolutional layers for extracting image features and predicting key point positions through a key point detection module.

[0015] The MRHKN architecture adds convolutional layers based on RKN for feature extraction and key point detection.

[0016] As a further improvement of the present invention, the MRHKN architecture uses mean square error as the loss function, and the optimization goal is to minimize the gap between the predicted key point position and the true position:

[0017] in, is the predicted key point location, is the actual key point position, and N is the number of key points.

[0018] As a further improvement of the present invention, the PnP algorithm estimates the camera's pose, i.e., the translation vector and the rotation matrix, through known 3D points and their corresponding 2D projection points:

[0019] Where P represents the camera's pose, R is the rotation matrix, t is the translation vector, Xi is the 3D point in the world coordinate system, xi is the corresponding 2D projection point, and π is the perspective projection function.

[0020] In a second aspect, the present invention provides a vision-based electric bus positioning system, comprising: An acquisition module is used to obtain a data set with charging station images and annotate them to obtain annotated charging station images; A labeling module, used for inputting the labeled charging station image into the neural network model for training, so as to obtain a neural network model for key point detection; A detection module for detecting key points on the charging station using a key point detection neural network model; A matching module is used to match the detected key points with the known charging station model to obtain a matched 3D-2D correspondence; The solution module is used to use the matched 3D-2D correspondence to solve the camera's pose through the PnP algorithm.

[0021] In a third aspect, the present invention provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the vision-based electric bus positioning method when executing the computer program.

[0022] In a fourth aspect, the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the vision-based electric bus positioning method is implemented.

[0023] In a fifth aspect, the present invention provides a computer program product, wherein the computer program product includes computer instructions, and the computer instructions instruct a computer to execute the vision-based electric bus positioning method.

[0024] The beneficial effects of the present invention compared with the prior art are as follows: The use of low-cost hardware devices, such as low-resolution cameras, reduces the overall cost of the system. Through efficient neural network models and geometric positioning algorithms, fast and accurate positioning is achieved, improving the efficiency of docking charging stations. Through automation technology, the reliance on driver experience is reduced, and the repetitive work and potential errors caused by manual operation are reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the embodiments of the present invention or the drawings of related technical solutions in the prior art are introduced below. It should be understood that the drawings introduced below are only for the convenience of clearly describing some embodiments of the technical solutions of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0026] Figure 1 A vision-based electric bus positioning method proposed by the present invention; Figure 2 Example images from the test dataset under different weather and lighting conditions Figure 3 The keypoint locations estimated for the charger pose; Figure 4 This is a schematic diagram of the RKN architecture; Figure 5 This is a schematic diagram of the MRHKN architecture; Figure 6 Cumulative distribution functions of 2D translation error (top) and orientation error (bottom) of the RKN and MRHKN methods, (a) for RKN; (b) for MRHKN method. DETAILED DESCRIPTION

[0027] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and are not to be construed as limitations of the present invention. For the step numbers in the following embodiments, they are only provided for the convenience of explanation, and the order between the steps is not limited in any way, and the execution order of each step in the embodiment can be adaptively adjusted according to the understanding of those skilled in the art.

[0028] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0029] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof. It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. The embodiments will be described in detail below in conjunction with the accompanying drawings.

[0030] The present invention aims to solve the problem of accurate positioning of electric buses during automatic docking with charging stations. In the prior art, there are certain limitations in using the global positioning system (GPS) and differential GPS (DGPS) for vehicle positioning. Especially in urban environments, GPS signals are easily blocked and interfered by high buildings, resulting in reduced positioning accuracy and failure to meet the high-precision requirements required for electric buses to dock with charging stations. In addition, the GPS system has an error of several meters in positioning, which is unacceptable for precise docking tasks.

[0031] See also Figure 1 The first object of the present invention is to propose a vision-based electric bus positioning method, comprising: A data set with images of charging stations is obtained, and the data set is annotated to obtain annotated images of charging stations; Input the labeled charging station image into the neural network model for training to obtain a neural network model for key point detection; Detect key points on the charging station using a key point detection neural network model; Match the detected key points with the known charging station model to obtain a matched 3D-2D correspondence; Using the matched 3D-2D correspondence, the camera's pose is solved using the PnP algorithm.

[0032] To solve these problems, this paper proposes a visual positioning system based on a monocular camera. The system uses deep learning technology to automatically detect key point features on the charging station through two neural network architectures: regression key point network (RKN) and maximum resolution heat map key point network (MRHKN). These key points are then used to calculate the precise 3D pose of the bus relative to the charging station through the perspective n-point algorithm.

[0033] The use of passive vision sensors eliminates the need to install expensive or active markers on charging stations, reducing deployment costs. In addition, the system has demonstrated good robustness through training and testing under diverse environmental conditions. Experimental results show that the system can provide accurate positioning information even at long distances and different angles, meeting the accuracy requirements required for electric buses to dock with charging stations.

[0034] In addition, existing vision-based positioning methods, such as visual odometry and SLAM, can provide relatively accurate positioning information, but they usually require a large number of salient features and may need to start from a known position, which may not be feasible in practical applications. Moreover, these methods perform poorly when dealing with small targets at a long distance, and when an electric bus approaches a charging station, the charging station may appear very small in the camera's field of view.

[0035] The method of the present invention is described in detail below in conjunction with specific embodiments: This paper proposes a novel visual localization method that uses a monocular camera system and deep learning techniques to improve localization accuracy. Specifically, this paper proposes two new neural network architectures - Regression Keypoint Network (RKN) and Maximum Resolution Heatmap Keypoint Network (MRHKN) to detect specific point features on the charging station. These networks are able to process high-resolution images and accurately estimate the position and posture of the bus relative to the charging station without the need for additional markers or equipment.

[0036] The present invention particularly emphasizes low cost and practicality, using passive visual sensors without installing any active components on the charging station, making the system easy to deploy and maintain. In addition, the robustness of the proposed method under different weather and lighting conditions, as well as the accuracy at long distances and different angles, are experimentally verified, thus demonstrating the potential of the method in practical applications. By solving these technical problems, the present invention provides an effective solution for the automatic docking of electric buses with charging stations, which helps to improve the operational efficiency and safety of electric buses.

[0037] The technical solution of the present invention achieves several key technical effects. First, the system provides high-precision positioning capabilities, which can accurately locate electric buses even at long distances and different angles. Second, the robustness of the system has been verified, and it can work stably under different weather and lighting conditions. In addition, the real-time performance of the system meets the needs of dynamic docking of electric buses. The system's ease of deployment and flexibility are also reflected, and it can adapt to different charging station structures and key point layouts.

[0038] Feature point detection and recognition specifically include: In order to achieve accurate positioning, the study adopted a vision-based approach to detect key points on the charging station. Specifically, the study proposed two new neural network architectures - RKN and MRHKN - for key point estimation. Both architectures are based on the design of convolutional neural network (CNN). Their performance was compared through a large number of experiments, and the MRHKN architecture was finally selected as the optimal solution. Figure 2 Example images from the test dataset under different weather and lighting conditions.

[0039] The neural network model is introduced as follows: RKN (Regular Keypoint Network): Based on regular convolutional layers and key point detection modules.

[0040] MRHKN (Modified Regular Keypoint Network): Improved on the basis of RKN, improving the accuracy of key point detection.

[0041] Training Dataset: The researchers collected a large dataset with images of charging stations and annotated them to train a neural network model for detecting key points. The annotations include the locations of key points on the charging station, which can be the four corners of the charging station head, natural corner points, or artificially marked points. Figure 3 Keypoint locations for the estimated charger pose.

[0042] The geometric method for estimating posture specifically includes: After detecting the key points, the next step is to use these points to estimate the pose of the bus. The study used the perspective-n-point (PnP) algorithm to achieve this. The PnP algorithm estimates the camera pose, i.e., the translation vector and rotation matrix, from known 3D points and their corresponding 2D projection points.

[0043]

[0044] Where P represents the camera's pose, R is the rotation matrix, t is the translation vector, Xi is the 3D point in the world coordinate system, xi is the corresponding 2D projection point, and π is the perspective projection function.

[0045] The specific steps of posture solving process are as follows: 1) Feature point detection: Use the MRHKN model to detect key points on the charging station.

[0046] The above MRHKN model refers to a specific feature point detection algorithm used to detect key points from an image. This model may be based on local feature descriptors such as SIFT, SURF or ORB to identify significant features in an image.

[0047] Specific process: In the image of the charging station, the MRHKN model will find and extract key points, which are usually corners, edges or other significant texture areas in the image. These points have good stability and repeatability under different viewing angles.

[0048] 2) Feature matching: Match the detected key points with the known charging station model.

[0049] The above feature matching can be performed in the following ways: Matching strategy: Once key points are detected in the charging station image, the next step is to match these points with key points in the known charging station model. This usually involves comparing feature descriptors to find the best correspondence.

[0050] Matching algorithm: You can use brute force matching, FLANN matching, or distance ratio-based matching algorithms to find the best matching pair. During the matching process, a threshold is usually set to filter out bad matches to improve the accuracy of the matching.

[0051] 3) PnP solution: Using the matched 3D-2D correspondence, the PnP algorithm is used to solve the camera's pose.

[0052] The above PnP solution process can be solved by the following method: PnP problem: The PnP (Perspective-n-Point) problem is a classic problem in computer vision. Its goal is to solve the camera pose (position and orientation) given a set of 3D points and their 2D projections on the image.

[0053] Algorithm selection: There are many PnP algorithms to choose from, such as direct linear transform (DLT), iterative closest point (ICP), P3P, AP3P, etc. Each algorithm has its applicable scenarios and advantages and disadvantages.

[0054] Solution process: In the PnP solution process, the algorithm uses the matched 3D-2D point pair relationship to estimate the camera's rotation matrix and translation vector through mathematical optimization methods. In this process, the least squares method is usually used to minimize the reprojection error, that is, to minimize the difference between the predicted 2D points and the actual observed 2D points.

[0055] The present invention also verifies the effectiveness of the method through experiments: the experiments cover scenes under different weather conditions and lighting conditions, including cloudy mornings, sunny noons, clear mornings, and foggy mornings. Through these experiments, the robustness and accuracy of the MRHKN architecture under various conditions are verified. Experiments were conducted using electric buses equipped with monocular cameras. The dataset includes images of charging stations taken under different weather and lighting conditions. Evaluation indicators include the accuracy of key point detection, the error of pose estimation, etc. The experimental results show that the MRHKN architecture outperforms the existing state-of-the-art detectors in key point detection and exhibits the best performance in estimating the position (translation and rotation) of the bus. Under different weather and lighting conditions, the MRHKN architecture can maintain high detection accuracy and pose estimation accuracy.

[0056] 1. Neural network architecture used in feature point detection: RKN architecture: It consists of multiple convolutional layers to extract image features and predict key point locations through a key point detection module. Figure 4 This is a schematic diagram of the RKN architecture; MRHKN architecture: Adds more convolutional layers based on RKN, further improving the ability of feature extraction and thus improving the accuracy of key point detection. Figure 5 This is a schematic diagram of the MRHKN architecture; Loss function: The mean square error (MSE) is used as the loss function, and the optimization goal is to minimize the gap between the predicted key point position and the true position.

[0057]

[0058] in, is the predicted key point location, is the actual key point position, and N is the number of key points.

[0059] The geometric method is described as follows: PnP algorithm: Estimate the camera's pose by solving the PnP problem. Use the EPnP (EfficientPerspective-n-Point) algorithm to accelerate the solution process. The EPnP algorithm simplifies the solution process by decomposing the problem into multiple sub-problems, thereby improving computational efficiency.

[0060] The present invention also evaluates the key point detection accuracy: evaluating the gap between the detected key point position and the true position.

[0061]

[0062] Pose estimation error: evaluates the difference between the estimated camera pose and the true pose.

[0063]

[0064] in, and are the estimated translation vector and rotation matrix, and t and R are the true values.

[0065] The vision-based positioning method proposed in this study can accurately locate the electric bus, thereby helping the electric bus to dock accurately at the charging station. This method combines the advantages of deep learning and geometric methods, uses a neural network model for key point detection, and uses the PnP algorithm for pose estimation. Experimental results show that this method can maintain high detection accuracy and pose estimation accuracy under different weather and lighting conditions, and has high practical value. Future research can further explore how to apply this method in more complex environments and optimize the algorithm to improve real-time performance.

[0066] Figure 6 Cumulative distribution function diagrams of 2D translation error (top) and orientation error (bottom) of RKN and MRHKN methods, (a) for RKN; (b) for MRHKN method, revealing the superiority of MRHKN method.

[0067] By using a low-cost monocular camera system instead of expensive multi-camera or LiDAR devices, the hardware cost of the system is greatly reduced. The study found that the required positioning accuracy can be achieved using a lower-resolution camera (such as a 4MP camera), thus reducing the need for high-resolution, high-cost cameras. By using a small number of passive markers instead of a large number of markers at the charging station, the cost of installation and maintenance is reduced. This method improves the speed and accuracy of electric buses docking charging stations through efficient neural network models and geometric positioning algorithms. Specifically: The proposed MRHKN architecture performs well in key point detection and can quickly and accurately detect key points on the charging station. The perspective-n-point (PnP) algorithm is used for pose estimation, which can complete the conversion from image to real-world coordinates in a short time, thus achieving fast positioning.

[0068] The traditional manual docking method requires the driver to have rich experience and skills, and each docking requires repeating the same high-precision operation. This method reduces the high dependence on the driver through automation technology, which is manifested in: Automated docking: Automatically identify and locate charging stations through the visual system, reducing the manual operations required by the driver and reducing the uncertainty and errors caused by human factors.

[0069] Driver Assistance System: The proposed ADAS (Advanced Driver Assistance System) can help the driver complete the docking task more easily and reduce the driver's workload.

[0070] The second object of the present invention is to provide a vision-based electric bus positioning system, comprising: An acquisition module is used to obtain a data set with charging station images and annotate them to obtain annotated charging station images; A labeling module, used for inputting the labeled charging station image into the neural network model for training, so as to obtain a neural network model for key point detection; A detection module for detecting key points on the charging station using a key point detection neural network model; A matching module is used to match the detected key points with the known charging station model to obtain a matched 3D-2D correspondence; The solution module is used to use the matched 3D-2D correspondence to solve the camera's pose through the PnP algorithm.

[0071] The third object of an embodiment of the present invention is to provide an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the vision-based electric bus positioning method when executing the computer program.

[0072] A fourth objective of an embodiment of the present invention is to provide a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the vision-based electric bus positioning method is implemented.

[0073] A fifth objective of an embodiment of the present invention is to provide a computer program product, wherein the computer program product comprises computer instructions, wherein the computer instructions instruct a computer to execute the vision-based electric bus positioning method.

[0074] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0075] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0076] The present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0077] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0078] Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0079] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A vision-based electric bus positioning method, characterized in that: include: A data set with images of charging stations is obtained, and the data set is annotated to obtain annotated images of charging stations; Input the labeled charging station image into the neural network model for training to obtain a neural network model for key point detection; Detect key points on the charging station using a key point detection neural network model; Match the detected key points with the known charging station model to obtain a matched 3D-2D correspondence; Using the matched 3D-2D correspondence, the camera's pose is solved using the PnP algorithm.

2. The electric bus positioning method based on vision according to claim 1 is characterized in that: The annotation includes key point positions on the charging station, specifically four corner points, natural corner points or artificial marking points at the head of the charging station.

3. The electric bus positioning method based on vision according to claim 1 is characterized in that: The neural network models are RKN architecture and MRHKN architecture.

4. The electric bus positioning method based on vision according to claim 3 is characterized in that: The RKN architecture consists of multiple convolutional layers to extract image features and predict key point locations through a key point detection module; The MRHKN architecture adds convolutional layers based on RKN for feature extraction and key point detection.

5. The electric bus positioning method based on vision according to claim 1 is characterized in that: The MRHKN architecture uses mean square error as the loss function, and the optimization goal is to minimize the difference between the predicted keypoint positions and the true positions: in, is the predicted key point location, is the actual key point position, and N is the number of key points.

6. The electric bus positioning method based on vision according to claim 1 is characterized in that: The PnP algorithm estimates the camera's pose, i.e., the translation vector and rotation matrix, through known 3D points and their corresponding 2D projection points: Where P represents the camera's pose, R is the rotation matrix, t is the translation vector, Xi is the 3D point in the world coordinate system, xi is the corresponding 2D projection point, and π is the perspective projection function.

7. A vision-based electric bus positioning system, characterized in that: include: An acquisition module is used to obtain a data set with charging station images and annotate them to obtain annotated charging station images; A labeling module, used for inputting the labeled charging station image into the neural network model for training, so as to obtain a neural network model for key point detection; A detection module for detecting key points on the charging station using a key point detection neural network model; A matching module is used to match the detected key points with the known charging station model to obtain a matched 3D-2D correspondence; The solution module is used to use the matched 3D-2D correspondence to solve the camera's pose through the PnP algorithm.

8. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the vision-based electric bus positioning method according to any one of claims 1 to 6 when executing the computer program.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the vision-based electric bus positioning method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising computer instructions, characterized in that: The computer instructions instruct the computer to execute the vision-based electric bus positioning method described in any one of claims 1-6.

Images