Railway overhead line system maintenance car auxiliary parking distance measurement method based on target detection

Abstract

The invention relates to a target detection-based auxiliary parking distance measurement method for a railway overhead line system maintenance car. The method comprises the steps that firstly, parking lot scene depth image data are acquired and preprocessed; then, training a target detection model by adopting the preprocessed parking lot scene depth image data, and outputting bounding box information of a target object after training; then, determining an initial center point coordinate based on the bounding box information of the target object, and determining an initial depth based on the parking lot scene depth image data; then, correcting the initial center point coordinate and the initial depth by adopting a space-time joint filtering algorithm, and determining the center point coordinate and the target depth of the target object; and finally, determining a target parking distance based on the coordinate of the central point of the target object and the target depth in combination with a depth image pixel value. The problems of single-frame data jitter and complex background interference are effectively solved, and the operation safety and parking efficiency of the maintenance car in a real operation scene are remarkably improved.

Description

Technical Field

[0001] The present application relates to the technical field of machine vision, and particularly to an auxiliary parking ranging method for a railway catenary maintenance vehicle based on object detection. Background Art

[0002] With the continuous expansion of the railway network and the increasing complexity of maintenance work, maintenance vehicles play a crucial role in the maintenance and repair of railway lines. Maintenance vehicles are usually equipped with multiple devices and tools to ensure the efficient execution of catenary maintenance tasks. However, the current operation of parking maintenance vehicles mostly relies on manual monitoring and guidance. Especially during the parking process of maintenance vehicles, maintenance personnel need to stand above the maintenance vehicle platform, observe with the naked eye and communicate with the driver using a walkie-talkie to ensure the precise docking of the vehicle body with the maintenance area. This operation method not only has low efficiency but also has certain safety hazards. Especially in complex environments and adverse weather conditions, problems such as unclear line of sight and poor communication are likely to occur.

[0003] Existing parking assistance technologies mainly include ultrasonic ranging, lidar, and vision-based ranging systems. Among them, ultrasonic and lidar ranging have a certain degree of accuracy, but are often limited in complex scenarios (such as occlusion, light changes, reflections, etc.). Vision-based ranging methods, especially technologies that combine object detection algorithms with depth cameras, have received extensive attention in recent years. Object detection algorithms based on neural networks can identify and locate objects in real time, and have great advantages in detection accuracy and efficiency. However, in the actual parking scenario of railway maintenance vehicles, most existing vision-based ranging systems can only be applied in static environments and have poor adaptability to occlusion and dynamic environments, so they cannot completely replace manual monitoring.

[0004] Therefore, in related technologies, there is an urgent need for a method that can improve the accuracy, adaptability, and robustness of the parking ranging method. Summary of the Invention

[0005] Based on this, in view of the above technical problems, it is necessary to provide an auxiliary parking ranging method for a railway catenary maintenance vehicle based on object detection that can improve the accuracy, adaptability, and robustness of the parking ranging method.

[0006] In a first aspect, the present application provides an auxiliary parking ranging method for a railway catenary maintenance vehicle based on object detection. The method includes:

[0007] Obtain the depth image data of the parking scene and perform preprocessing;

[0008] Use the preprocessed depth image data of the parking scene to train the object detection model, and output the object boundary box information after training;

[0009] Determine the initial center point coordinates based on the target object bounding box information, and determine the initial depth based on the depth image data of the parking scene;

[0010] Use a spatio-temporal joint filtering algorithm to correct the initial center point coordinates and the initial depth, and determine the target object center point coordinates and the target depth;

[0011] Determine the target parking distance based on the target object center point coordinates and the target depth in combination with the depth image pixel values.

[0012] Optionally, in an embodiment of the present application, the preprocessing includes denoising, light compensation, edge enhancement, and annotation.

[0013] Optionally, in an embodiment of the present application, the training of the target detection model using the depth image data of the parking scene after preprocessing includes:

[0014] Input the depth image data of the parking scene after preprocessing into the initial target detection model to obtain the initial object bounding box information;

[0015] Compare the initial object bounding box information with the annotated object bounding box information, adjust the model parameters according to the comparison result, and determine the optimal target detection model when the accuracy rate reaches the preset standard.

[0016] Optionally, in an embodiment of the present application, the use of the spatio-temporal joint filtering algorithm to correct the initial center point coordinates and the initial depth, and determine the target object center point coordinates and the target depth includes:

[0017] Use the continuity of the target motion trajectory to correct the instantaneous jitter;

[0018] Use the statistical characteristics of the depth distribution in the target neighborhood to suppress outliers.

[0019] Optionally, in an embodiment of the present application, the method further includes:

[0020] Restrict the depth value sampling area based on the initial center point coordinates.

[0021] In a second aspect, the present application also provides a railway catenary inspection vehicle auxiliary parking ranging device based on target detection. The device includes:

[0022] An image data acquisition and processing module, configured to acquire depth image data of a parking scene and perform preprocessing;

[0023] A target detection module, configured to train a target detection model using the depth image data of the parking scene after preprocessing, and output target object bounding box information after training;

[0024] The central point depth initial calculation module is used to determine the initial central point coordinates based on the target object bounding box information and determine the initial depth based on the depth image data of the parking scenario;

[0025] The correction module is used to correct the initial central point coordinates and the initial depth by using a spatio-temporal joint filtering algorithm to determine the target object central point coordinates and the target depth;

[0026] The parking distance determination module is used to determine the target parking distance based on the target object central point coordinates and the target depth in combination with the depth image pixel values.

[0027] In a third aspect, the present application also provides a computer device. The computer device includes a memory and a processor. The memory stores a computer program, and the processor executes the steps of the methods described in the above respective embodiments.

[0028] In a fourth aspect, the present application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, and when the computer program is executed by a processor, the steps of the methods described in the above respective embodiments are implemented.

[0029] For the above-mentioned railway catenary maintenance vehicle auxiliary parking ranging method based on target detection, first, obtain the depth image data of the parking scenario and perform preprocessing; then, use the preprocessed depth image data of the parking scenario to train the target detection model, and after training, output the target object bounding box information; then, determine the initial central point coordinates based on the target object bounding box information and determine the initial depth based on the depth image data of the parking scenario; then, correct the initial central point coordinates and the initial depth by using a spatio-temporal joint filtering algorithm to determine the target object central point coordinates and the target depth; finally, determine the target parking distance based on the target object central point coordinates and the target depth in combination with the depth image pixel values. That is to say, aiming at the technical defect that the traditional ranging method is vulnerable to illumination changes, occlusion and reflection interference during the parking process of the maintenance vehicle, the advantages of target detection and depth camera technology are innovatively integrated, and a precise ranging method with environmental adaptability is proposed. By combining the three-dimensional data obtained in real time by the depth camera with the target center coordinates accurately located by the target detection algorithm, supplemented by multi-frame data fusion and spatio-temporal filtering processing technology, the problems of single-frame data jitter and complex background interference are effectively overcome. This solution not only ensures the positioning accuracy through the stable recognition ability of the target detection algorithm under occlusion conditions, but also uses the three-dimensional perception characteristics of the depth camera to provide a reliable benchmark for ranging. Finally, the system has both millimeter-level ranging accuracy and strong anti-interference ability in a dynamic and complex environment, significantly improving the operation safety and docking efficiency of the maintenance vehicle in the actual operation scenario. Description of the Drawings

[0030] Figure 1Schematic flowchart of the auxiliary parking ranging method for a railway catenary maintenance vehicle based on object detection in an embodiment;

[0031] Figure 2 Schematic wiring diagram of the ranging system in an embodiment;

[0032] Figure 3 Structural block diagram of the auxiliary parking ranging device for a railway catenary maintenance vehicle based on object detection in an embodiment;

[0033] Figure 4 Internal structure diagram of a computer device in an embodiment. Detailed implementation manners

[0034] In order to make the objectives, technical solutions and advantages of the present application clearer and more understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0035] In one embodiment, as Figure 1 shown, an auxiliary parking ranging method for a railway catenary maintenance vehicle based on object detection is provided, including the following steps:

[0036] S101: Obtain the depth image data of the parking scene and perform preprocessing.

[0037] In the embodiments of the present application, in an actual railway maintenance scene or a laboratory scene, a ZED depth camera is used to collect the image data required for object detection, which is divided into a training set and a validation set. Ensure that the collected image data covers different lighting conditions, such as strong sunlight during the day, cloudy days, dusk, night, fluctuating light intensity, etc., to improve the robustness and adaptability of the object detection model. The image data should include scenes with different parking distances, involving catenary-related components (such as insulators) and other possible objects to be detected. Considering the diversity of the railway maintenance environment, the collected images should include complex backgrounds (such as sundries, background buildings, light reflections, etc.) to train the model's robustness in complex environments.

[0038] In one embodiment of the present application, the preprocessing includes denoising, light compensation, edge enhancement, and annotation.

[0039] In an embodiment of the present application, for each frame of the acquired image, image preprocessing is performed, such as denoising, illumination compensation, edge enhancement, etc. Then, the labelimg software of the annotation tool is used, and according to the requirements of object detection, the bounding boxes of the target objects in the images of the training set are accurately annotated. The annotation of each target object needs to include the target category (such as an insulator) and the coordinates of the bounding box of the target object (the upper left and lower right coordinates), ensuring that all the acquired images are annotated according to a unified annotation specification to avoid missing or incorrect annotations. For each image, the annotation should include the position of the bounding box of each target object and its corresponding class label, providing accurate training data for subsequent object detection and distance estimation algorithms.

[0040] S103: Use the depth image data of the parking scene after preprocessing to train the object detection model, and output the bounding box information of the target object after training.

[0041] In the embodiment of the present application, the YOLOV5 object detection model is trained with the acquired training set, so that the model can identify the target objects (such as insulators) in the maintenance vehicle parking scene, real-time locate the targets in the image, calculate the bounding boxes (bounding boxes) of the targets and output them.

[0042] Specifically, in an embodiment of the present application, the training of the object detection model using the depth image data of the parking scene after preprocessing includes:

[0043] S201: Input the depth image data of the parking scene after preprocessing into the initial object detection model to obtain the initial object bounding box information.

[0044] S203: Compare the initial object bounding box information with the annotated object bounding box information, adjust the model parameters according to the comparison results, and determine the optimal object detection model when the accuracy reaches the preset standard.

[0045] In an embodiment of the present application, the image data in the training set is input into the initial YOLOV5 object detection model to learn how to identify target objects (such as insulators of the catenary) from the images. The YOLOV5 algorithm extracts features in the image through a convolutional neural network (CNN), and then predicts the position and category of the bounding box of the target object. Through the image data in the validation set, the performance of the object detection model is tested. In this process, according to the preset evaluation criteria (such as accuracy, recall rate, IoU, etc.), the effect of the model in the actual scene is evaluated. Through the feedback of the validation set, the model parameters can be adjusted and optimized to make the performance of the model reach the optimal in the actual application.

[0046] S105: Determine the initial center point coordinates based on the target object bounding box information, and determine the initial depth based on the parking scene depth image data.

[0047] In the embodiments of the present application, the initial center point coordinates of the bounding box of the target object output by the target detection model are calculated by the bounding box coordinates (xyxy) of the target object. Where xyxy are four values, which are the coordinates of the upper left corner and the lower right corner respectively: (x 1 , y 1 , x 2 , y 2 ). These coordinates are normalized coordinates relative to the input image, and the range is [0, 1]. After being scaled by GN (the width and height of the input image), the coordinates in the actual image are obtained. The coordinates of the initial center point mid_pos are calculated by the following formula.

[0048]

[0049] Among them, mid_pos x represents the horizontal midpoint of the target box, and mid_pos y represents the vertical midpoint of the target box.

[0050] At the same time, the initial depth is determined based on the parking scene depth image data obtained by the ZED depth camera. The depth image data provides the distance from the camera for each pixel point, that is, the depth value. The depth value corresponding to the corresponding position is obtained based on the initial center point coordinates, that is:

[0051] d = depth_value[y, x]

[0052] Among them, y and x are the calculated center point coordinates, and depth_value is the pixel value of the depth image data.

[0053] S107: Use the spatio-temporal joint filtering algorithm to correct the initial center point coordinates and the initial depth, and determine the target object center point coordinates and the target depth.

[0054] In the embodiments of the present application, by combining the temporal motion trajectory of the target detection box with the spatial neighborhood depth information, the ranging error is dynamically corrected. In the filtering process, a target motion detection model is introduced, and the speed information of the maintenance vehicle is used to predict the target position and match it with the depth data in real time to reduce jitter and solve the problem of ranging lag caused by the rapid movement of the target in a dynamic scene.

[0055] In an embodiment of the present application, the step of using the spatio-temporal joint filtering algorithm to correct the initial center point coordinates and the initial depth, and determine the target object center point coordinates and the target depth includes:

[0056] S301: Use the continuity of the target motion trajectory to correct the instantaneous jitter.

[0057] S303: Suppress outliers by using the statistical characteristics of the depth distribution in the target neighborhood.

[0058] In an embodiment of the present application, time - dimension filtering (Kalman filter prediction) is first performed. Let the initial center - point coordinates of the target in the image be (x t , y t ), and the corresponding depth value be d t . The state equation and observation equation of the Kalman filter are defined as follows:

[0059] State vector:

[0060]

[0061] where v x,t , v y,t are the velocity components of the target in the image, and is the center - point coordinates of the target directly detected by the target detection algorithm in the current frame.

[0062] State prediction equation:

[0063] X t|t-1 = FX t-1 + w t ,

[0064] where Δt is the time interval between frames, and w t ~ N(0, Q) is the process - noise covariance matrix.

[0065] Observation update equation:

[0066] Z t = HX t|t-1 + v t ,

[0067] where v t ~ N(0, R) is the observation - noise covariance matrix.

[0068] Through the prediction and update steps of the Kalman filter, the smoothed target position is obtained to suppress instantaneous jitter.

[0069] After that, spatial - dimension filtering (local consistency constraint) is performed. For the depth data within the target detection box, a local neighborhood Ω (an N×N window centered at the center point ) is defined, and the statistic of the depth value is calculated:

[0070] Depth mean:

[0071]

[0072] Among them, d i,j is the depth value at the pixel position (i, j) in the depth image.

[0073] Depth variance:

[0074]

[0075] If the observed depth of the current frame satisfies (k is a threshold, usually taken as 2 to 3), it is determined as an outlier and replaced with instead.

[0076] The final depth estimate is the weighted fusion of time prediction and spatial correction:

[0077]

[0078] Among them, is the depth value predicted by the Kalman filter (based on the corresponding position), is the local depth mean value after spatial filtering

[0079] The adaptive weight α is designed as:

[0080]

[0081] Among them, ‖v t ‖ is the magnitude of the target motion speed, v th is the speed threshold, and β is the adjustment parameter.

[0082] When the target moves fast (‖v t ‖ > v th ), α → 1, relying on time prediction;

[0083] When the target is stationary or moving slowly (‖v t ‖ < v th ), α → 0, relying on spatial correction.

[0084] S109: Determine the target parking distance based on the center point coordinates of the target object and the target depth in combination with the depth image pixel values.

[0085] In the embodiments of the present application, since each pixel of the camera corresponds to a certain physical depth, the pixel value in the depth map can be directly used as the distance from the target object to the camera. Let the depth value extracted from the depth map be d, then the physical distance from the target object to the camera is D, and the unit is usually meters (m). The unit of the depth map is millimeters and needs to be converted to meters:

[0086]

[0087] The formula for calculating the final target parking distance based on the coordinates of the center point of the target object is as follows:

[0088]

[0089] In an embodiment of the present application, the method further includes:

[0090] Restricting the depth value sampling area based on the initial center point coordinates.

[0091] In an embodiment of the present application, in order to improve the accuracy of distance measurement, taking mid_pos as the center, sampling the depth values within this range, and obtaining the final sampling area by taking the minimum value:

[0092] min_val = min(|x 2 - x 1 |, |y 2 - y 1 |)

[0093] min_val represents the size of the border and is used to determine the width of the sampling area. This step is to improve the stability and accuracy of distance measurement, especially for small targets or blurred areas in the image.

[0094] In an embodiment of the present application, as Figure 2 shown, the camera line is connected to the industrial computer of the lifting platform. The industrial computer then leads out a line and introduces it into the driver's cab through the reserved wiring hole on the vehicle roof, and is connected to the display device. The measured parking distance is displayed using a simple UI window. In the center of the window, the text "Distance: XX.X meters" is displayed in red. The distance changes in real time according to the driving of the maintenance vehicle. The UI window is displayed on a 7-inch IPS small display screen installed in the driver's cab, which is used to feedback the distance measurement information to the maintenance vehicle driver in real time. Using a 7-inch IPS small display screen to transmit the distance measurement information simplifies information transmission, improves decision-making efficiency, ensures that the driver can make accurate and rapid parking decisions in a complex working environment, and at the same time optimizes system performance and reduces calculation and bandwidth burdens.

[0095] In the above-mentioned auxiliary parking ranging method for railway catenary maintenance vehicles based on object detection, first, depth image data of the parking scene is acquired and preprocessed; then, the preprocessed depth image data of the parking scene is used to train an object detection model, and after training, the object bounding box information is output; then, based on the object bounding box information, the initial center point coordinates are determined, and based on the depth image data of the parking scene, the initial depth is determined; then, the spatio-temporal joint filtering algorithm is used to correct the initial center point coordinates and the initial depth to determine the object center point coordinates and the target depth; finally, based on the object center point coordinates and the target depth, combined with the depth image pixel values, the target parking distance is determined. That is to say, aiming at the technical defects of traditional ranging methods being vulnerable to illumination changes, occlusion, and reflection interference during the parking process of maintenance vehicles, the advantages of object detection and depth camera technology are innovatively integrated, and an accurate ranging method with environmental adaptability is proposed. By combining the three-dimensional data obtained in real time by the depth camera with the target center coordinates accurately located by the object detection algorithm, supplemented by multi-frame data fusion and spatio-temporal filtering processing technology, the problems of single-frame data jitter and complex background interference are effectively overcome. This solution not only ensures the positioning accuracy through the stable recognition ability of the object detection algorithm under occlusion conditions, but also uses the three-dimensional perception characteristics of the depth camera to provide a reliable benchmark for ranging. Finally, the system has both millimeter-level ranging accuracy and strong anti-interference ability in a dynamic and complex environment, significantly improving the operation safety and docking efficiency of the maintenance vehicle in the actual operation scenario.

[0096] It should be understood that although the steps in the flowcharts involved in the above-mentioned embodiments are shown in sequence according to the arrows, these steps do not necessarily need to be executed in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments may include multiple steps or multiple stages. These steps or stages do not necessarily need to be executed at the same time, but can be executed at different times. The execution order of these steps or stages does not necessarily need to be sequential, but can be executed alternately or in turn with at least a part of other steps or steps in other steps.

[0097] Based on the same inventive concept, an embodiment of the present application further provides a target detection-based auxiliary parking ranging device for a railway catenary maintenance vehicle for implementing the above-mentioned target detection-based auxiliary parking ranging method for a railway catenary maintenance vehicle. The implementation solutions provided by this device to solve problems are similar to those recorded in the above method. Therefore, the specific limitations in one or more of the following embodiments of the target detection-based auxiliary parking ranging device for a railway catenary maintenance vehicle can refer to the limitations on the target detection-based auxiliary parking ranging method for a railway catenary maintenance vehicle in the above text, and will not be repeated here.

[0098] In one embodiment, as Figure 3 shown, a target detection-based auxiliary parking ranging device 300 for a railway catenary maintenance vehicle is provided, including: an image data acquisition and processing module 301, a target detection module 303, a center point depth initial calculation module 305, a correction module 307, and a parking distance determination module 309, where:

[0099] The image data acquisition and processing module 301 is configured to acquire depth image data of the parking scene and perform preprocessing.

[0100] The target detection module 303 is configured to train a target detection model using the preprocessed depth image data of the parking scene, and output target object bounding box information after training.

[0101] The center point depth initial calculation module 305 is configured to determine an initial center point coordinate based on the target object bounding box information, and determine an initial depth based on the depth image data of the parking scene.

[0102] The correction module 307 is configured to correct the initial center point coordinate and the initial depth using a spatio-temporal joint filtering algorithm to determine the target object center point coordinate and the target depth.

[0103] The parking distance determination module 309 is configured to determine the target parking distance based on the target object center point coordinate and the target depth in combination with the depth image pixel value.

[0104] In one embodiment of the present application, the preprocessing includes denoising, light compensation, edge enhancement, and annotation.

[0105] In one embodiment of the present application, the target detection module is further configured to:

[0106] Input the preprocessed depth image data of the parking scene into an initial target detection model to obtain initial object bounding box information;

[0107] Compare the initial object bounding box information with the annotated object bounding box information, adjust the model parameters according to the comparison result, and determine the optimal target detection model when the accuracy reaches a preset standard.

[0108] In one embodiment of the present application, the correction module is further configured to:

[0109] Adopt the continuity of the target motion trajectory to correct the instantaneous jitter;

[0110] Adopt the statistical characteristics of the depth distribution in the target neighborhood to suppress outliers.

[0111] In one embodiment of the present application, the method further includes:

[0112] Restrict the depth value sampling area based on the initial center point coordinates.

[0113] Each module in the above-mentioned railway catenary maintenance vehicle auxiliary parking ranging device based on target detection can be implemented in whole or in part by software, hardware, and their combination. Each of the above modules can be embedded in the processor of the computer device in hardware form or be independent of it, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each of the above modules.

[0114] In one embodiment, a computer device is provided. The computer device can be a terminal, and its internal structure diagram can be as Figure 4 shown. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner. The wireless manner can be implemented through WIFI, a mobile cellular network, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it realizes a method for auxiliary parking ranging of a railway catenary maintenance vehicle based on target detection. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the shell of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0115] Those skilled in the art can understand that Figure 4 the structure shown in

[0116] In one embodiment, a computer device is provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in the above method embodiments are implemented.

[0117] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0118] In one embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0119] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0120] Those of ordinary skill in the art can understand that all or part of the processes in the above-described method embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the method embodiments as described above. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., without limitation.

[0121] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0122] The above-described embodiments only represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A method for assisted parking distance measurement of a railway overhead line maintenance vehicle based on target detection, characterized in that: The method comprises: Acquire parking scene depth image data and perform preprocessing; The pre-processed parking scene depth image data is used to train the target detection model, and the target object bounding box information is output after training; Determine the initial center point coordinates based on the target object bounding box information, and determine the initial depth based on the parking scene depth image data; Using a spatiotemporal joint filtering algorithm to correct the initial center point coordinates and initial depth, and determine the center point coordinates and target depth of the target object; The target parking distance is determined based on the target object center point coordinates and the target depth combined with the depth map pixel value.

2. The method for assisted parking distance measurement of a railway overhead line maintenance vehicle based on target detection according to claim 1 is characterized in that: The preprocessing includes denoising, illumination compensation, edge enhancement and labeling.

3. The railway overhead line maintenance vehicle auxiliary parking distance measurement method based on target detection according to claim 1 is characterized in that: The method of training the target detection model using the pre-processed parking scene depth image data includes: Input the preprocessed parking scene depth image data into the initial target detection model to obtain the initial object bounding box information; The initial object bounding box information is compared with the labeled object bounding box information, and the model parameters are adjusted according to the comparison result. When the accuracy reaches a preset standard, the optimal target detection model is determined.

4. The railway overhead line maintenance vehicle auxiliary parking distance measurement method based on target detection according to claim 1 is characterized in that: The adopting of a spatiotemporal joint filtering algorithm to correct the initial center point coordinates and the initial depth to determine the center point coordinates and the target depth of the target object comprises: Adopt the continuity of the target motion trajectory to correct the instantaneous jitter; The statistical characteristics of the depth distribution of the target neighborhood are used to suppress outliers.

5. The railway overhead line maintenance vehicle auxiliary parking distance measurement method based on target detection according to claim 1 is characterized in that: The method further comprises: The depth value sampling area is limited based on the initial center point coordinates.

6. A railway overhead line maintenance vehicle auxiliary parking distance measurement device based on target detection, characterized in that: The device comprises: An image data acquisition and processing module is used to obtain parking scene depth image data and perform preprocessing; The target detection module is used to train the target detection model using the pre-processed parking scene depth image data, and output the target object bounding box information after training; A center point depth initial calculation module, used to determine the initial center point coordinates based on the target object boundary box information, and determine the initial depth based on the parking scene depth image data; A correction module, used to correct the initial center point coordinates and initial depth using a spatiotemporal joint filtering algorithm to determine the center point coordinates and target depth of the target object; The parking distance determination module is used to determine the target parking distance based on the center point coordinates of the target object and the target depth in combination with the depth map pixel value.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images