High-precision railway map generation method and system based on one-way trajectory correction

Abstract

The invention provides a high-precision railway map generation method and system based on one-way trajectory correction, and relates to the technical field of heavy haul railway transportation, and the method comprises the steps: collecting a continuous image sequence and spatial attitude data of a current train in a driving process at the same time; determining pose change data of the current train based on visual feature points in the continuous image sequence; performing synchronous positioning and map construction based on the pose change data and the spatial pose data to generate an initial electronic map; and based on mileage information and image positions corresponding to preset markers in the continuous image sequence, the mileage and the track of the current train in the initial electronic map are corrected, and a corresponding electronic map of the current train in the driving process is generated. According to the method and system provided by the invention, the generation efficiency and accuracy of the railway electronic map are improved.

Description

Technical Field

[0001] The present application relates to the technical field of heavy-haul railway transportation, and in particular to a method and system for generating a high-precision railway map based on one-way track correction. Background Art

[0002] Heavy-haul rail transport plays a vital role worldwide. The safe operation, efficient planning and refined maintenance of railway systems increasingly rely on accurate electronic maps. Therefore, the development of high-precision railway electronic maps faces huge technical demands and development potential.

[0003] However, traditional railway surveying methods show obvious limitations when dealing with long distances and complex terrain. For example, mobile surveying vehicles and drone aerial surveys, although they can provide a certain degree of accuracy, require a lot of time and resources, especially in long-distance railways with complex terrain. In addition, these methods have limitations in data fusion and processing, especially when multi-source heterogeneous data is fused, it is difficult to achieve a unified expression of data integration from different sensors, which affects the accuracy of the map and the robustness of the system.

[0004] Therefore, how to improve the efficiency and accuracy of railway electronic map generation has become a technical problem that needs to be solved urgently in the industry. Summary of the invention

[0005] The present application provides a high-precision railway map generation method and system based on one-way track correction, which are used to solve the technical problem of how to improve the generation efficiency and accuracy of railway electronic maps.

[0006] The present application provides a method for generating a high-precision railway map based on one-way trajectory correction, comprising: At the same time, the continuous image sequence and spatial posture data of the current train during its travel are collected; Determining the posture change data of the current train based on the visual feature points in the continuous image sequence; Performing synchronous positioning and map construction based on the position change data and the spatial posture data to generate an initial electronic map; Based on the mileage information and image positions corresponding to the preset markers in the continuous image sequence, the mileage and track of the current train in the initial electronic map are corrected to generate an electronic map corresponding to the current train during its travel.

[0007] In some embodiments, after simultaneously acquiring the continuous image sequence and spatial posture data of the current train during its travel, the method further includes: Preprocessing the continuous image sequence; the preprocessing includes at least one of distortion correction, denoising and image enhancement; Filtering binocular images in the continuous image sequence based on timestamps of the left-eye image and the right-eye image in the continuous image sequence; The spatial posture data is synchronized and calibrated based on the timestamps of the binocular images in the continuous image sequence.

[0008] In some embodiments, the synchronous positioning and map construction based on the posture change data and the spatial posture data to generate an initial electronic map includes: Determining a state vector based on actual measurement values ​​of each sensor; the sensor includes a binocular camera for collecting the continuous image sequence and an inertial measurement unit for collecting the spatial posture data; Determine a state estimation cost function based on actual measurement values ​​and expected calculated values ​​of each sensor at each moment, and a covariance matrix of the actual measurement values; Optimizing and solving the state estimation cost function to obtain an optimal estimation value of the state vector of the current train during its travel; Based on the optimal estimated value of the state vector, synchronous positioning and map construction are performed to generate an initial electronic map.

[0009] In some embodiments, based on the mileage information and image positions corresponding to the preset markers in the continuous image sequence, the mileage and track of the current train in the initial electronic map are corrected to generate the electronic map corresponding to the current train during the travel process, including: Performing landmark recognition on the continuous image sequence to determine the target type of each preset landmark in the continuous image sequence; Based on the target type of each preset marker, determine the mileage information corresponding to each preset marker; The mileage information corresponding to each preset marker is compared with the measured mileage of the current train, and the mileage of the current train in the initial electronic map is corrected based on the mileage deviation between the mileage information and the measured mileage.

[0010] In some embodiments, based on the mileage information and image positions corresponding to the preset markers in the continuous image sequence, the mileage and track of the current train in the initial electronic map are corrected to generate the electronic map corresponding to the current train during the travel process, including: Performing landmark recognition on the continuous image sequence to determine the image position of each preset landmark in the continuous image sequence; Based on the image position and the spatial coordinate position of each preset marker, and the parameters of the binocular camera that collects the continuous image sequence, determine the actual position information of the binocular camera; Based on the actual position information of the binocular camera, the track of the current train in the initial electronic map is corrected.

[0011] In some embodiments, the performing of landmark recognition on the continuous image sequence includes: Inputting the continuous image sequence into a landmark recognition model to obtain the target type, image position and confidence of each preset landmark output by the landmark recognition model; The marker recognition model is obtained by training based on sample images corresponding to each preset marker.

[0012] In some embodiments, after the mileage and track of the current train in the initial electronic map are corrected based on the mileage information and image positions corresponding to the preset markers in the continuous image sequence to generate the electronic map corresponding to the current train during the travel, the method further includes: Determine the latitude and longitude coordinates of the starting point of the current train in the electronic map; Determine the longitude and latitude coordinates of each track point based on the longitude and latitude coordinates of the starting point and the relative displacement data of each track point of the current train; Based on the longitude and latitude coordinates, the track point number, the road section name, the road section type and the road section length of each track point, a key point table, a road section table and a road section details table corresponding to the electronic map are generated.

[0013] The present application provides a high-precision railway map generation system based on one-way track correction, comprising: A data acquisition module is used to simultaneously acquire the continuous image sequence and spatial posture data of the current train during its travel; A feature extraction module, used for determining the posture change data of the current train based on the visual feature points in the continuous image sequence; A synchronous mapping module, used for performing synchronous positioning and mapping based on the posture change data and the spatial posture data to generate an initial electronic map; The map correction module is used to correct the mileage and track of the current train in the initial electronic map based on the mileage information and image positions corresponding to the preset markers in the continuous image sequence, and generate an electronic map corresponding to the current train during its travel.

[0014] The present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method for generating a high-precision railway map based on one-way trajectory correction is implemented.

[0015] The present application provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for generating a high-precision railway map based on one-way trajectory correction is implemented.

[0016] The high-precision railway map generation method and system based on one-way trajectory correction provided by the present application simultaneously collect continuous image sequences and spatial posture data of the current train during travel; determine the posture change data of the current train based on the visual feature points in the continuous image sequence; perform synchronous positioning and map construction based on the posture change data and the spatial posture data to generate an initial electronic map; based on the mileage information and image positions corresponding to the preset markers in the continuous image sequence, correct the mileage and trajectory of the current train in the initial electronic map to generate the electronic map corresponding to the current train during travel; due to the combination of the continuous image sequence collected by the binocular camera and the spatial posture data collected by the inertial measurement unit, high-precision positioning and map construction in the railway environment are achieved, and the positioning and mapping problems under high-speed movement of the train can be handled more accurately, and the stability and accuracy of the electronic map construction are improved; due to the use of the mileage information and image positions of the preset markers in the railway environment, the mileage and trajectory of the current train in the initial electronic map are corrected, which effectively reduces the cumulative error in long-distance travel, and is particularly suitable for one-way long-distance railway scenes where traditional closed-loop detection cannot be performed, and the generation efficiency and accuracy of railway electronic maps are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0018] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 It is a flow chart of a method for generating a high-precision railway map based on one-way trajectory correction provided by the present application.

[0020] Figure 2 It is one of the structural schematic diagrams of the high-precision railway map generation system based on one-way trajectory correction provided by this application.

[0021] Figure 3 This is an architecture diagram of a high-precision railway map generation system based on one-way trajectory correction provided by this application.

[0022] Figure 4 This is the second structural schematic diagram of the high-precision railway map generation system based on one-way trajectory correction provided by this application.

[0023] Figure 5 It is a method flow chart of a high-precision railway map generation system based on one-way trajectory correction provided by the present application.

[0024] Figure 6 It is a structural schematic diagram of the electronic device provided by this application. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0026] It should be noted that the terms "first", "second" etc. in the present application are used to distinguish similar objects, and need not be used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable in appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, the process, method, system, product or equipment comprising a series of steps or units or modules need not be limited to those steps or units or modules clearly listed, but may include other steps or units or modules that are not clearly listed or inherent to these processes, methods, products or equipment.

[0027] When building high-precision maps, relevant technologies are divided into two categories: traditional methods such as mobile mapping vehicles or drone aerial surveys, and non-traditional methods such as remote sensing images and trajectory data production. These methods involve multiple key technologies such as multi-source heterogeneous data fusion, simultaneous localization and mapping (SLAM), artificial intelligence large models, knowledge graphs, etc.

[0028] Although mobile mapping vehicles and drones can provide a certain degree of accuracy, they require a lot of time and resources, especially in long-distance railways with complex terrain. In addition, these methods have limitations in data fusion and processing, especially when fusion of multi-source heterogeneous data, it is difficult to achieve a unified expression of data integration from different sensors, thus affecting the accuracy of the map and the robustness of the system.

[0029] Although SLAM technology has made progress in the fields of autonomous driving and robot navigation, its application in railway map construction is still full of challenges. For example, visual SLAM does not work well in railway tunnel environments with insufficient lighting or missing features. Although laser SLAM can generate high-resolution point cloud maps, it is expensive and has a slow processing speed. In addition, existing SLAM technology often has inter-frame registration problems when processing high-speed motion scenes, affecting the stability and accuracy of the system.

[0030] At the same time, although the large AI model can significantly improve environmental perception capabilities, how to efficiently integrate and process large-scale data in the railway environment and how to optimize the model for complex and changing railway scenarios are still major challenges. In addition, although the knowledge graph has rich semantic information and decision support potential, how to effectively use this knowledge and its dynamic update and maintenance in railway map construction is also one of the challenges facing the technology.

[0031] In order to solve the shortcomings of related technologies, Figure 1 is a flow chart of a method for generating a high-precision railway map based on one-way track correction provided by the present application, such as Figure 1 As shown, the method includes step 110 , step 120 , step 130 and step 140 .

[0032] Step 110: simultaneously collect the continuous image sequence and spatial posture data of the current train during its travel.

[0033] Specifically, the execution subject of the method provided in the embodiment of the present application is a high-precision railway map generation system. The system can be implemented by software, such as a high-precision railway map generation program running on a computer or server; it can also be implemented by hardware, such as a mobile terminal, computer or server that executes the high-precision railway map generation method.

[0034] The application scenario of the method provided in the embodiment of the present application is to set a binocular camera and an inertial measurement unit (IMU) on the current train, and the setting position can be the head of the train. The current train is a train that constructs a railway electronic map in real time during driving.

[0035] The continuous image sequence is a series of images that are continuous in time, which are generated by the binocular camera capturing the environment along the way in real time during the train's running. It can be understood that the images collected by the binocular camera include left-eye images and right-eye images.

[0036] The spatial attitude data is the three-axis velocity and angular velocity collected by the inertial measurement unit during the train's travel. It can be understood that the inertial measurement unit contains three single-axis accelerometers and three single-axis gyroscopes. The accelerometer detects the acceleration signal of the object in the independent three-axis of the carrier coordinate system, and the gyroscope detects the angular velocity signal of the carrier relative to the navigation coordinate system, measuring the angular velocity and acceleration of the object in three-dimensional space, and using this to calculate the object's attitude.

[0037] During the current train movement, continuous image sequences and spatial posture data are collected simultaneously.

[0038] Step 120: Determine the posture change data of the current train based on the visual feature points in the continuous image sequence.

[0039] Specifically, the ORB (Oriented FAST and Rotated BRIEF) algorithm is an algorithm that combines FAST corner detection and BRIEF feature descriptor. The ORB algorithm can be used to extract visual feature points from continuous image sequences, and the extracted visual feature points can be used for matching. By comparing the position changes of these visual feature points, the camera's posture change data in space can be inferred. Posture includes position and attitude. Since the camera is fixed on the head of the train, the camera's posture change data is actually the posture change data of the current train.

[0040] Step 130: Perform synchronous positioning and map construction based on the position change data and the spatial posture data to generate an initial electronic map.

[0041] Specifically, the spatial attitude data is the data collected by the inertial measurement unit, and the motion trajectory of the camera, including position and attitude (orientation), can be obtained through integral calculation.

[0042] Through optimization algorithms, the posture change data (visual information) and spatial posture data (inertial information) can be effectively integrated. Through the SLAM algorithm in related technologies, the accuracy and robustness of positioning and mapping can be improved, thereby outputting the initial electronic map.

[0043] Step 140: Based on the mileage information and image positions corresponding to the preset markers in the continuous image sequence, the mileage and track of the current train in the initial electronic map are corrected to generate an electronic map corresponding to the current train during its travel.

[0044] Specifically, considering that trains usually travel long distances in one direction and do not have a closed path, the closed-loop detection in related technologies cannot be effectively implemented.

[0045] The method provided in the embodiment of the present application can identify preset markers in a continuous image sequence. The preset markers may include road markers such as kilometer markers, half-kilometer markers, hundred-meter markers, tunnel entrances, and tunnel exits. These road markers are usually used to indicate fixed distances in a railway environment, have corresponding mileage information, and can be used as reference points to assist the positioning process. In addition, the image position of the preset marker can be obtained through continuous image sequence recognition, and the camera's posture can be calculated through the image position and the physical size of the preset marker, thereby obtaining the train's posture.

[0046] The mileage of the current train in the initial electronic map can be corrected through the mileage information corresponding to the preset markers, and the trajectory of the current train in the initial electronic map can be corrected through the image position corresponding to the preset markers, so as to effectively eliminate the accumulated errors generated by the long-distance travel of the train, and finally generate the electronic map corresponding to the current train during the travel process.

[0047] The high-precision railway map generation method based on one-way trajectory correction provided by the embodiment of the present application simultaneously collects the continuous image sequence and spatial posture data of the current train during the travel process; determines the posture change data of the current train based on the visual feature points in the continuous image sequence; performs synchronous positioning and map construction based on the posture change data and the spatial posture data to generate an initial electronic map; based on the mileage information and image positions corresponding to the preset markers in the continuous image sequence, corrects the mileage and trajectory of the current train in the initial electronic map to generate the electronic map corresponding to the current train during the travel process; due to the combination of the continuous image sequence collected by the binocular camera and the spatial posture data collected by the inertial measurement unit, high-precision positioning and map construction in the railway environment are achieved, and the positioning and mapping problems under high-speed movement of the train can be handled more accurately, and the stability and accuracy of the electronic map construction are improved; due to the use of the mileage information and image positions of the preset markers in the railway environment, the mileage and trajectory of the current train in the initial electronic map are corrected, which effectively reduces the cumulative error in long-distance travel, is particularly suitable for one-way long-distance railway scenes where traditional closed-loop detection cannot be performed, and improves the generation efficiency and accuracy of railway electronic maps.

[0048] It should be noted that each implementation method of the present application can be freely combined, the order can be changed, or it can be executed separately, and does not need to rely on or depend on a fixed execution order.

[0049] In some embodiments, after simultaneously acquiring the continuous image sequence and the spatial posture data of the current train during its travel, the method further includes: Preprocessing the continuous image sequence; the preprocessing includes at least one of distortion correction, denoising and image enhancement; Based on the timestamps of the left eye image and the right eye image in the continuous image sequence, the binocular images in the continuous image sequence are filtered; The spatial posture data is synchronized and calibrated based on the timestamps of the binocular images in the continuous image sequence.

[0050] Specifically, the continuous image sequence captured by the binocular camera can be preprocessed, for example, image correction can be performed using preset distortion parameters, and then denoising and image enhancement can be performed.

[0051] The images in the continuous image sequence include the left eye image and the right eye image collected in pairs. The timestamps of the left eye image and the right eye image collected in pairs are compared. If the difference between the timestamps of the two exceeds the preset difference, it indicates that the two images are not collected synchronously and can be discarded at the same time; if the difference between the timestamps of the two is less than or equal to the preset difference, it indicates that the two images are collected synchronously and can be retained.

[0052] For the spatial attitude data collected by the inertial measurement unit, the timestamp of the spatial attitude data can be compared with the timestamp of the binocular image in the continuous image sequence for synchronization and calibration, so that the binocular image and the spatial attitude data can be aligned on the time axis. At the same time, the random error in the configuration file is used to initialize the zero bias of the acceleration and angular velocity in the spatial attitude data, and the processed data is used for subsequent use.

[0053] The high-precision railway map generation method based on one-way trajectory correction provided in the embodiment of the present application ensures the accuracy and consistency of the data by preprocessing and correcting the images taken by the binocular camera, and synchronizing and calibrating the spatial posture data, thereby improving the accuracy of the railway electronic map.

[0054] In some embodiments, synchronous positioning and map construction are performed based on the position change data and the spatial posture data to generate an initial electronic map, including: The state vector is determined based on the actual measurement values ​​of each sensor; the sensors include a binocular camera for collecting continuous image sequences and an inertial measurement unit for collecting spatial posture data; Determine a state estimation cost function based on the actual measurement values ​​and expected calculated values ​​of each sensor at each moment, and the covariance matrix of the actual measurement values; Optimize and solve the state estimation cost function to obtain the optimal estimated value of the state vector of the current train during its travel; Based on the optimal estimate of the state vector, simultaneous positioning and map construction are performed to generate an initial electronic map.

[0055] Specifically, when the train runs into a tunnel section, the number of extracted feature points drops sharply due to poor lighting conditions, which may lead to positioning failure. In this environment with limited visual information, the key to optimizing the SLAM algorithm is to increase the system's reliance on non-visual sensor data and enhance the algorithm's efficiency in utilizing existing features.

[0056] The embodiment of the present application provides a SLAM optimization method in a visual information-constrained environment, which supports combining visual data and IMU data, treating the measurement value of each sensor as a factor, and summarizing the factors sharing the same state variables into a state vector, and realizing accurate estimation of the train state through state optimization. The method includes: Step 1: System status definition

[0057] The state vector is determined according to the actual measurement values ​​of each sensor; the sensors include a binocular camera for collecting continuous image sequences and an inertial measurement unit for collecting spatial posture data.

[0058] During system (train) localization, the system aims to estimate its exact position in three-dimensional space ( ) and posture ( In addition, it is also necessary to estimate the depth of the visual landmarks observed by the camera ( ), which is the straight-line distance from the camera to each detected feature point, and the motion variable generated by the IMU: speed ( ), acceleration bias and gyro bias .

[0059] State Vector Usually includes dynamic parameters such as position, attitude, speed, etc., which are used to describe the instantaneous state of the system in the environment:

[0060] For vision systems, index Represents a specific step in the time series, each step corresponds to a set of data collected from sensors that are used to calculate and update the position and attitude of the train. Represents the depth value of each feature point in the initial frame (the initial image in the continuous image sequence) ; For an inertial measurement unit (inertial sensor), the state vector This includes the velocity vector and two sensor bias parameters and .

[0061] Step 2: State estimation cost function definition and weight adjustment

[0062] The state estimation problem is constructed as a maximum likelihood estimation (MLE) problem in the framework of , and the goal is to solve it by optimizing the joint probability distribution of the posture over a period of time. In this process, it is assumed that all measurements are independent and the uncertainty of the measurement follows a Gaussian distribution. Based on these assumptions, the problem can be further transformed into a nonlinear least squares problem, namely, bundle adjustment (BA).

[0063] The state estimation cost function can be determined based on the actual measured values ​​and expected calculated values ​​of each sensor at each moment, as well as the covariance matrix of the actual measured values, as follows:

[0064] in, Represents the state vector The optimal estimate of , which aims to minimize the error of the entire system; is a collection of measurements from cameras, IMUs, and other sensors; is the observation vector, which contains the actual measurement data (actual measurement values) obtained from the sensor; It is a time index, used to mark a specific point in time of a state or measurement; is a set of measurements A specific measurement value in Indicated in At this moment, the system obtains observation data from each sensor; Represents observation data The covariance matrix of , which quantifies the uncertainty of the observations; is the sensor model, express At time t, the expected value of each sensor (expected calculated value) calculated from the current attitude and position of the system.

[0065] When the train enters the tunnel, the ambient light will decrease rapidly, resulting in a sharp decrease in the number of feature points that can be extracted. At this time, the reliability of visual information will be reduced. Therefore, it is necessary to appropriately increase the weight of IMU information in the SLAM process to reduce the positioning error caused by insufficient visual information. The two states of the train entering and exiting the tunnel can be judged by environmental feature detection. In the cost function, the measurement values ​​of each sensor appear in the form of residuals, and their importance is determined by the corresponding covariance matrix. In order to increase the influence of IMU data and reduce the influence of camera data under specific conditions, these covariance matrices can be adjusted. Specifically, by reducing the covariance matrix of IMU measurements, the uncertainty weight of IMU data is actually reduced, giving it greater influence in the optimization process. Conversely, increasing the covariance matrix of camera measurements will increase the uncertainty weight of camera data, correspondingly reducing its relative importance in state estimation.

[0066] In mathematical expression, the cost function can be expressed as:

[0067] in, Represents the transpose operation of a matrix; and Represent the measurement sets of the camera and IMU sensor respectively; and is the covariance matrix of the corresponding sensor measurement. By adjusting these covariance matrices through the ratio of extracted feature points to the preset feature points to be extracted, the influence of different sensor data in the optimization process can be effectively controlled.

[0068] Step 3: Sensor Factor Definition

[0069] The Camera Factor represents the stereo camera. Corner features are detected in each camera frame and tracked between frames (single images) using the KLT (Kanade-Lucas-Tomasi) tracker. Based on feature associations, the Camera Factor for each feature in each frame is constructed.

[0070]

[0071] in, Indicated in Features in the image at this moment The observed pixel coordinates of Indicates based on system status ,exist Features in the image at this moment The predicted pixel coordinates of It's in time (i.e., a specific starting frame); and It's in time The camera rotation matrix of a subsequent frame. These matrices describe the rotation from the world coordinate system to the camera coordinate system. It's in time The camera position; and It's in time The camera position; express The distance from the camera to the current feature point at any moment; is the camera projection function, which is used to map a three-dimensional point in the camera coordinate system to a two-dimensional point in the image plane; and Represents time and The transformation matrix of the camera's posture (rotation and translation) at this time. is the external transformation matrix from the system center to the camera center, which is pre-calibrated offline; , is The features in the image at the moment The horizontal and vertical coordinates of the initial observed pixel, is The horizontal and vertical coordinates of the same feature pixels observed at all times.

[0072] The IMU Factor uses a known IMU pre-integration algorithm that assumes the additive noise of the acceleration and gyroscope measurements is Gaussian white noise.

[0073]

[0074] in, Indicates time arrive The IMU measurement vector between them includes the integrated relative position, velocity, and orientation changes. Indicates from the state The obtained IMU prediction vector. , , Respectively represent the relative position, velocity, and rotation measurements integrated from the IMU. express The rotation matrix of the system at time instant. and They are and The position of the system in three-dimensional space at that moment. Represents the acceleration due to gravity. and Respectively and The speed of the system at all times. express and The time interval between. and Respectively and The deviation of the acceleration reading at that moment. and Respectively and The deviation of the gyroscope reading at that moment.

[0075] Step 4: Optimization solution

[0076] The cost function is solved by the Gauss-Newton method. The cost function is relative to the state The initial guess value of After linearization, the following results are obtained:

[0077] in, is the state vector The optimization process tries to find an increment that minimizes the total residual sum of squares. It's in time and measurement The residual vector under represents the predicted state and actual observation The error between . is the residual About Status The Jacobian matrix of , that is, the partial derivatives of the residual with respect to the state vector. It is measured In time The covariance matrix of is used to weight the residuals during the minimization process to ensure that observation uncertainties are properly accounted for.

[0078] After linear approximation, the cost function is obtained The closed form solution of , the solution obtained is as follows:

[0079] in, Represents the Hessian Matrix, which is the matrix of all Jacobian matrices A weighted sum of the second-order partial derivatives with respect to the forecast errors. This vector is a cumulative term that includes the weighted residuals over all time steps and measurements.

[0080] Then, the solution is Used to update the current status ,pass (in is an addition operation on the manifold used for rotations) to update. This process is iterated multiple times until convergence.

[0081] Step 5: Marginalization

[0082] In order to control the computational complexity, historical measurements are converted into prior terms through marginalization, preserving past information without causing information loss. The marginalization operation is completed using Schur complement, which converts the information about the marginalized state into a prior term:

[0083] , , , These symbols represent the information matrix different parts of which: is the sub-block of the information matrix associated with the state variables to be marginalized. and ( The transpose of ) represents the cross-information matrix part between the marginalized state variables and the retained state variables. is the sub-block of the information matrix associated with the retained state variables. Represents the increment of the retained state variable, which is the variable that needs to be solved during the optimization process. , Represents information vector different parts of is the part related to the state variable to be marginalized, and This is the part related to retaining state variables. It is the new information matrix obtained through Schur complement, which is used to update the optimization problem of retaining the state. is the updated information vector for state-preserving optimization problems.

[0084] In this way, a new prior on the remaining state is obtained , which contains the information of the marginalized state without any information loss. The system maintains a fixed number of spatial camera frames. When a new keyframe (filtered image) arrives, the visual and inertial factors associated with the first frame are marginalized.

[0085] After obtaining prior information about the current state using Bayes' rule, the posterior is calculated as the product of likelihood and prior: Then, state estimation becomes a maximum a posteriori (MAP) problem. To time At the moment The previous state is marginalized and converted into a prior. Therefore, the MAP problem is written as:

[0086] +( )

[0087] Finally, the final state can be obtained by using a nonlinear optimizer, and the optimal estimated value of the state vector of the current train during its travel can be obtained. Based on the optimal estimated value of the state vector, synchronous positioning and map construction can be performed to generate an initial electronic map.

[0088] The high-precision railway map generation method based on one-way trajectory correction provided by the embodiment of the present application realizes accurate estimation of the train state by fusing binocular video and IMU data in an optimization framework. This process involves state definition, cost function establishment, sensor factor construction, optimization process design, and marginalization processing, which ensures the efficiency and flexibility of the system. Through appropriate marginalization and optimization strategies, the present application can effectively retain the support capabilities for different sensor configurations. When the train is in an environment with insufficient lighting, the use of inertial measurement unit (IMU) data can significantly improve the robustness of the SLAM system. The acceleration and angular velocity information provided by the IMU can be used to estimate the state of motion in the short term and help bridge the lack of visual information. By accurately pre-integrating the IMU data and closely integrating it with the visual data, the continuity and accuracy of positioning can be maintained when visual features are scarce.

[0089] In some embodiments, based on the mileage information and image positions corresponding to the preset markers in the continuous image sequence, the mileage and track of the current train in the initial electronic map are corrected to generate the electronic map corresponding to the current train during the driving process, including: Performing landmark recognition on the continuous image sequence to determine the target type of each preset landmark in the continuous image sequence; Based on the target type of each preset marker, determine the mileage information corresponding to each preset marker; The mileage information corresponding to each preset marker is compared with the measured mileage of the current train, and the mileage of the current train in the initial electronic map is corrected based on the mileage deviation between the mileage information and the measured mileage.

[0090] Specifically, when a train travels long distances, it will generate cumulative errors, including mileage errors and track errors, etc. Preset markers such as kilometer markers, half-kilometer markers, and hundred-meter markers on the railway (each kilometer marker represents a fixed point along the line, usually at intervals of 1 kilometer, and the same applies to half-kilometer markers and hundred-meter markers) and the train's own odometer can be used to perform accurate mileage correction.

[0091] When a train passes a kilometer mark and is successfully identified, the system compares the distance measured by the odometer with the actual fixed distance (usually 1 kilometer). If there is a deviation, that is, the distance measured by the odometer does not match the actual distance between the kilometer marks, the system will adjust the real-time position of the train according to the deviation value to correct the mileage error. The steps of mileage correction include: Step 1: Marker Identification

[0092] An integrated neural network model (such as a YOLO model) can be used to identify landmarks in a continuous image sequence to determine the target type of each preset landmark in the continuous image sequence. The target type may include kilometer landmarks, half-kilometer landmarks, hundred-meter landmarks, etc.

[0093] Step 2: Get the actual location

[0094] Each preset marker can be regarded as a fixed point. For each identified fixed point, the system can obtain its corresponding mileage information, that is, the actual distance along the route. For example, a kilometer mark represents a distance of 1 kilometer. The actual position of the fixed point is .

[0095] Step 3: Mileage data comparison

[0096] The train's odometer provides the current measured mileage, that is, the distance The system compares the distance measured by the odometer with the actual distance of the identified fixed point. When there are fixed points, the distance measured by the odometer is .

[0097] Step 4: Mileage Correction

[0098] If the train passes through When there are fixed points, the distance measured by the odometer is , and the actual distance of the fixed point is (e.g. 1 km). Mileage deviation It can be calculated by the following formula: .

[0099] If there is a deviation , then the real-time position of the train needs to be corrected. Suppose the position of the train before adjustment is , the adjusted position is The correction can be done using the following formula: .

[0100] in, Based on the deviation The calculated correction vector. The specific calculation method is to normalize the deviation vector and multiply it by a correction factor, that is: ×Correction factor. The correction factor can be set according to actual conditions.

[0101] Take advantage of new real-time locations , the system can optimize the historical trajectory. Suppose the position sequence in the historical trajectory is , , …, , the system can be based on the deviation of each position Make adjustments.

[0102] Use the weighted average algorithm to optimize the historical trajectory and reduce the impact of cumulative errors. For example, the mathematical expression of using the weighted average method to optimize the historical trajectory can be:

[0103] in, is the optimized position, is the deviation after averaging at each key position.

[0104] The high-precision railway map generation method based on one-way trajectory correction provided in the embodiment of the present application accurately corrects the train position by real-time identification of preset landmark features at fixed distances and combining odometer data, thereby effectively eliminating the cumulative errors generated during long-distance driving.

[0105] In some embodiments, based on the mileage information and image positions corresponding to the preset markers in the continuous image sequence, the mileage and track of the current train in the initial electronic map are corrected to generate the electronic map corresponding to the current train during the driving process, including: Performing landmark recognition on the continuous image sequence to determine the image position of each preset landmark in the continuous image sequence; Based on the image position and spatial coordinate position of each preset marker and the parameters of the binocular camera that collects the continuous image sequence, the actual position information of the binocular camera is determined; Based on the actual position information of the binocular camera, the trajectory of the current train in the initial electronic map is corrected.

[0106] Specifically, the trajectory error can also be corrected by identifying preset markers. Similar to mileage correction, when identifying each preset marker, for each identified fixed point, the system will obtain its position information in the image and the corresponding category label. Suppose the image coordinates of the identified fixed point are ,in Indicates the serial number of the identified fixed point.

[0107] Once the preset marker is identified, the system will obtain the position of the preset marker in the image and calculate the camera's pose based on the physical size and known position of the preset marker.

[0108] Assume that the spatial coordinate position (world coordinate) of the preset marker is , the camera's intrinsic parameter matrix is , the projection of the logo in the image is According to the relationship between camera projection and world coordinates, the camera's pose can be calculated using the following formula:

[0109] in, and Represent the rotation matrix and translation vector of the camera respectively, is the camera’s intrinsic parameter matrix, It is the sign in Projection in frame image.

[0110] Based on the recognized landmark pose, the system can correct the actual pose of the camera. and is the pose before correction, and is the corrected posture. The posture correction can be performed by the following formula:

[0111] in, and is the corrected rotation and translation, is the rotation and translation obtained from landmark recognition, There may be deviations.

[0112] After the posture correction, the system needs to update the trajectory. For the keyframe (filtered image) To keyframe (filtered image) The relative transformation of is the relative transformation after correction.

[0113] The trajectory update can be performed using the following formula:

[0114] here, yes The inverse transformation of is the correction transformation obtained from landmark recognition.

[0115] Through the above steps, the system can correct the trajectory and posture by identifying specific environmental signs without traditional closed-loop detection. This method can effectively reduce the cumulative error and improve the positioning accuracy and robustness of the system.

[0116] The high-precision railway map generation method based on one-way trajectory correction provided in the embodiment of the present application corrects the trajectory of the current train in the initial electronic map according to the actual posture information of the binocular camera, thereby effectively eliminating the accumulated errors generated during long-distance travel.

[0117] In some embodiments, performing landmark recognition on a continuous image sequence includes: Inputting a continuous image sequence into a landmark recognition model, and obtaining a target type, an image position, and a confidence level of each preset landmark output by the landmark recognition model; Among them, the landmark recognition model is trained based on sample images corresponding to each preset landmark.

[0118] Specifically, mileage correction and trajectory correction mainly rely on identifying preset markers along the railway, such as kilometer markers, half-kilometer markers, hundred-meter markers, tunnel entrances and other fixed structures, which become highly reliable positioning reference points due to their uniqueness and equidistant nature.

[0119] A landmark recognition model can be constructed through a neural network model, and the target type, image location, and confidence level of each preset landmark can be obtained through the landmark recognition model. The model type of the neural network model can be selected as the YOLO (You Only Look Once) model. The target types of preset landmarks include kilometer marks, half-kilometer marks, hundred-meter marks, and tunnel entrance marks. The specific steps include: Step 1: Data collection and annotation

[0120] Collect sample images corresponding to each preset landmark. For example, collect a large number of pictures or video data of railway signs, especially images containing kilometer signs, half-kilometer signs, hundred-meter signs, and tunnel entrances. Then, manually annotate these data, indicate the type and location of each feature, and prepare a training set for the YOLO model.

[0121] To ensure the generalization of the model, the collected image and video data should cover different road sections. Therefore, when drawing the map, the weather conditions should be good and the visibility should be high. In addition, in order to improve the recognition accuracy, high-resolution images should be used and taken from multiple angles and positions, including bird's-eye view, side view, and front view.

[0122] During the data preprocessing phase, strict data cleaning is performed to remove blurry, overexposed, and poor quality data. At the same time, for ease of processing, the images are cropped to remove irrelevant backgrounds and scaled to a uniform size. In addition, data enhancement techniques such as rotation and flipping are used to further improve the model's adaptability to different situations.

[0123] The identification tags are divided into four categories: 0 km mark, 1.5 km mark, 2.100 m mark, and 3.tunnel entrance.

[0124] After labeling, the complete dataset is divided into training set, validation set, and test set in a ratio of 8:1:1 for use in subsequent model training and evaluation.

[0125] Through the above steps, a high-quality railway landscape dataset can be constructed, laying a solid foundation for the training of the YOLO model.

[0126] Step 2: Train the YOLO target detection model

[0127] The YOLO object detection model is trained using annotated datasets to enable the model to identify and locate railway landmarks in images. As an end-to-end deep learning model, the YOLO model can predict the location and category of objects in an image through a single forward propagation, which gives it significant advantages in speed and accuracy, making it particularly suitable for application scenarios that require real-time processing.

[0128] Step 3: Identify preset landmarks in real time

[0129] The trained YOLO model is integrated into the train's SLAM system. During the train's travel, the camera on the train captures the environment along the way in real time and generates a continuous sequence of images. These image data are used as input to the YOLO model after image preprocessing (including image resizing and normalizing pixel values). The railway's iconic features are identified in real time.

[0130] For each frame of the image, the YOLO model obtains the prediction result through forward propagation calculation. The output of the YOLO model can be expressed as a three-dimensional tensor of S×S×(5+C), where S is the size of the grid, C is the number of categories, and 5 represents the four coordinate values ​​and confidence score of each bounding box. Assuming that the size of the input image is W×H, W is the width and H is the height, the YOLO model divides it into S grids, and each grid is responsible for predicting the target whose center point falls within the grid. For each grid, the model predicts B bounding boxes and the corresponding confidence score and category probability for each bounding box.

[0131] For each predicted bounding box, the YOLO model outputs its coordinates in the image (x, y, w, h), where x and y are the image center coordinates of the preset marker, w is the image width of the preset marker, and h is the image height of the preset marker. These coordinates are normalized relative to the size of the grid. These coordinates can be converted to actual coordinates in the image by the following formula: real coordinates = normalized coordinates × (x / S, H / S). At the same time, the model also outputs the confidence score of each bounding box. and class probabilities The confidence score indicates how confident the model is that the predicted bounding box contains the object, and the class probability indicates how confident the model is that the object belongs to a specific class.

[0132] For each detected object, the SLAM system will receive the following information: Target type: Based on highest class probability Determine the target category.

[0133] Image location: The location of the object in the image according to the transformed bounding box coordinates (x, y, w, h).

[0134] Confidence: The confidence score of the model for the target detection , which is used to evaluate the reliability of the test results.

[0135] The SLAM system integrates the feature information returned by the YOLO model into the train positioning and map building process. It assists the train in determining its position in the railway network by identifying railway landmarks and updating the map information along the railway.

[0136] The high-precision railway map generation method based on one-way trajectory correction provided in the embodiment of the present application uses a deep learning model (such as YOLO) to identify specific landmark features in the railway environment (such as kilometer markers, tunnel entrances, etc.), providing rich reference information for subsequent positioning and mapping.

[0137] In some embodiments, based on the mileage information and image positions corresponding to the preset markers in the continuous image sequence, after correcting the mileage and track of the current train in the initial electronic map and generating the electronic map corresponding to the current train during the driving process, the method further includes: Determine the latitude and longitude coordinates of the starting point of the current train in the electronic map; Based on the latitude and longitude coordinates of the starting point and the relative displacement data of each track point of the current train, the latitude and longitude coordinates of each track point are determined; Based on the latitude and longitude coordinates, trajectory point number, section name, section type and section length of each track point, a key point table, section table and section detail table corresponding to the electronic map are generated.

[0138] Specifically, the generated electronic map can be post-processed to further improve and optimize the map information to output the final target format. The original output data of the system mainly includes information such as timestamp, three-axis relative displacement and system posture, but in order to build a detailed railway map, key geographic information such as longitude and latitude, kilometer mark, node type, etc. needs to be integrated.

[0139] The precise latitude and longitude coordinates of the starting point can be obtained through configuration files, etc. In order to obtain the most accurate coordinate data, RTK (Real-Time Kinematic) technology can be used for marking. Once the starting point is determined, the system will combine the relative displacement data in the x and y directions, and calculate the latitude and longitude coordinates of each subsequent track point through the integration method to ensure the continuity and accuracy of the map location information. In addition, the system can also identify and mark different types of nodes, such as starting points, end points, kilometer markers, half-kilometer markers, hundred-meter markers, tunnel entrances and tunnel exits. In this way, the map optimization and generation module can convert the original track data into a detailed and accurate railway map, and will output it in the target format for use. The target format can include key point tables, section tables, and section detail tables, as shown in Tables 1 to 3.

[0140] Table 1 Electronic map key point table

[0141] Table 2 Electronic map road section table

[0142] Table 3 Electronic map road section details

[0143] The high-precision railway map generation method based on one-way trajectory correction provided in the embodiment of the present application integrates key geographic information such as longitude and latitude, kilometer marks, node types, etc. in the generated electronic map, thereby improving the availability and accuracy of the railway electronic map.

[0144] The system provided in the embodiment of the present application is described below. The system described below and the method described above can be referenced to each other.

[0145] Figure 2 This is one of the structural schematic diagrams of the high-precision railway map generation system based on one-way trajectory correction provided by this application, such as Figure 2 As shown, the system includes: The data acquisition module 210 is used to simultaneously acquire the continuous image sequence and spatial posture data of the current train during its travel; A feature extraction module 220, for determining the posture change data of the current train based on the visual feature points in the continuous image sequence; A synchronous mapping module 230 is used to perform synchronous positioning and map construction based on the posture change data and the spatial posture data to generate an initial electronic map; The map correction module 240 is used to correct the mileage and track of the current train in the initial electronic map based on the mileage information and image positions corresponding to the preset markers in the continuous image sequence, and generate an electronic map corresponding to the current train during its travel.

[0146] The high-precision railway map generation system based on one-way trajectory correction provided by the embodiment of the present application simultaneously collects the continuous image sequence and spatial posture data of the current train during the travel process; determines the posture change data of the current train based on the visual feature points in the continuous image sequence; performs synchronous positioning and map construction based on the posture change data and the spatial posture data to generate an initial electronic map; based on the mileage information and image positions corresponding to the preset markers in the continuous image sequence, corrects the mileage and trajectory of the current train in the initial electronic map, and generates the electronic map corresponding to the current train during the travel process; due to the combination of the continuous image sequence collected by the binocular camera and the spatial posture data collected by the inertial measurement unit, high-precision positioning and map construction in the railway environment are achieved, and the positioning and mapping problems under high-speed movement of the train can be handled more accurately, and the stability and accuracy of the electronic map construction are improved; due to the use of the mileage information and image positions of the preset markers in the railway environment, the mileage and trajectory of the current train in the initial electronic map are corrected, and the accumulated error in long-distance travel is effectively reduced. It is particularly suitable for one-way long-distance railway scenes where traditional closed-loop detection cannot be performed, and the generation efficiency and accuracy of railway electronic maps are improved.

[0147] Figure 3 This is an architecture diagram of a high-precision railway map generation system based on one-way trajectory correction provided by this application. Figure 3As shown in the figure, the system uses multiple sensors to collect environmental information, obtains the forward serialized image of the train through the binocular camera, and uses the IMU on the camera to collect the three-axis angular velocity and linear acceleration of the train. For image information, after preprocessing, the system extracts the feature points of the left and right cameras, continuously tracks, and removes unqualified points. For IMU data, align gravity, obtain the initial posture of the IMU, and continuously update the posture using the median integral. After the tight coupling initialization of the binocular camera and the IMU is completed, the least squares problem is constructed using the prior information, the residuals of the left and right cameras, the IMU residuals, and the current environmental factors to solve the posture. When a milepost is detected, the system will optimize the historical posture, and the optimized results will be post-processed to form a complete map.

[0148] Figure 4 This is the second structural diagram of the high-precision railway map generation system based on one-way track correction provided by this application, such as Figure 4 As shown, the system can be divided into 6 modules. The specific modules and their functions are as follows: (1) Data acquisition module, installed at the head of the train, including a binocular camera and an IMU. During the train's travel, the binocular camera continuously acquires binocular video, and the IMU acquires the train's three-axis speed and angular velocity in real time, and transmits these data to the central processing unit.

[0149] (2) The data preprocessing module processes the binocular video data and IMU data respectively. For binocular images, the system will use the preset distortion parameters to perform image correction, and then perform denoising and image enhancement. At the same time, the timestamps of the left and right images are compared. If the difference between the two exceeds the threshold, the image is discarded. For IMU data, it is necessary to compare with the timestamp of the video data for synchronization and calibration. At the same time, the random error in the configuration file is used to initialize the zero bias of acceleration and angular velocity. The processed data is for subsequent use.

[0150] (3) The feature extraction and analysis module is mainly divided into two parts: target environment feature detection and specific image information enhancement. In terms of target environment feature detection, the system uses the YOLO pre-trained model to identify key landmarks with semantic information along the railway, such as kilometer markers, half-kilometer markers, hundred-meter markers, and tunnel entrances. Through the accurate detection of these features, the system can obtain accurate geographic location information, providing an important basis for subsequent map generation. In terms of specific image information enhancement, the system uses a series of image processing technologies to improve the quality of image data to ensure that the system can still operate stably in visually restricted environments. This includes contrast adjustment, sharpening, denoising, and other processing of the image to enhance the useful information in the image and suppress noise. Through these enhancement measures, the system can improve the stability of its own positioning and mapping. The processed feature information will be sent to the next module for further data analysis and map construction.

[0151] (4) The SLAM core processing module performs system positioning and mapping by fusing binocular video data and IMU data. By matching the visual features extracted from the binocular video data and comparing the position changes of these feature points, the camera's position change in space can be inferred. At the same time, the acceleration and angular velocity information provided by the IMU data is used to calculate the camera's motion trajectory through integration. These motion trajectories can be used to estimate the camera's position, including position and orientation. Through the optimization algorithm, the visual and inertial information are effectively integrated to improve the accuracy and robustness of positioning and mapping. Thus, a preliminary mapping result is output.

[0152] (5) The trajectory correction and error elimination module is an alternative to the traditional SLAM closed-loop detection method. Considering that trains usually travel long distances in one direction and do not have a closed path, traditional closed-loop detection cannot be effectively implemented. This system uses fixed-distance features in the railway environment, such as kilometer markers, half-kilometer markers, and hundred-meter markers as reference points to assist the positioning process. By matching these fixed-position features, the odometer error can be corrected and the historical posture can be corrected. This method overcomes the limitations of traditional SLAM closed-loop detection in long-distance one-way travel and improves the accuracy and reliability of the generated trajectory.

[0153] (6) The map optimization and generation module further improves and optimizes the map information by post-processing the generated map to output the final target format.

[0154] Figure 5 is a method flow chart of a high-precision railway map generation system based on one-way track correction provided by the present application, such as Figure 5As shown in the figure, the system integrates binocular camera and IMU data acquisition, preprocessing, feature extraction, SLAM processing, trajectory correction, map generation and optimization modules, and realizes the automatic processing of the entire process of railway electronic maps. Through the real-time data transmission of the data acquisition module and the precise processing of the preprocessing module, the system can effectively obtain key information during the train travel. The feature extraction and analysis module improves the robustness of positioning and environmental perception through binocular vision and environmental feature detection. The SLAM core processing module combines visual and inertial information for positioning and mapping, and the trajectory correction module uses fixed distance features to correct positioning errors. Finally, the map optimization of the global BA algorithm improves the accuracy and stability of the map. These works together constitute a complete system, which greatly improves the efficiency and accuracy of railway electronic map generation. The advantage of the system is that it makes full use of multi-sensor data, effectively copes with the challenges of long-distance one-way travel of trains, and provides reliable support for the safe operation of railways.

[0155] Figure 6 is a schematic diagram of the structure of the electronic device provided by this application, such as Figure 6 As shown, the electronic device may include: a processor (Processor) 610, a communication interface (Communications Interface) 620, a memory (Memory) 630 and a communication bus (Communications Bus) 640, wherein the processor 610, the communication interface 620, and the memory 630 communicate with each other through the communication bus 640. The processor 610 may call the logic command in the memory 630 to execute the method described in the above embodiment, for example: At the same time, the continuous image sequence and spatial posture data of the current train during its travel are collected; the posture change data of the current train is determined based on the visual feature points in the continuous image sequence; synchronous positioning and map construction are performed based on the posture change data and the spatial posture data to generate an initial electronic map; based on the mileage information and image positions corresponding to the preset landmarks in the continuous image sequence, the mileage and trajectory of the current train in the initial electronic map are corrected to generate an electronic map corresponding to the current train during its travel.

[0156] In addition, the logical commands in the above-mentioned memory can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several commands to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0157] The processor in the electronic device provided in the embodiment of the present application can call the logic instructions in the memory to implement the above method. Its specific implementation method is consistent with the implementation method of the aforementioned method and can achieve the same beneficial effects, which will not be repeated here.

[0158] An embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the methods provided in the above embodiments are implemented.

[0159] Its specific implementation is consistent with the aforementioned method implementation and can achieve the same beneficial effects, so it will not be repeated here.

[0160] An embodiment of the present application provides a computer program product, including a computer program, which implements the above method when executed by a processor.

[0161] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0162] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0163] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A high-precision railway map generation method based on one-way trajectory correction, characterized in that: include: At the same time, the continuous image sequence and spatial posture data of the current train during its travel are collected; Determining the posture change data of the current train based on the visual feature points in the continuous image sequence; Performing synchronous positioning and map construction based on the position change data and the spatial posture data to generate an initial electronic map; Based on the mileage information and image positions corresponding to the preset markers in the continuous image sequence, the mileage and track of the current train in the initial electronic map are corrected to generate an electronic map corresponding to the current train during its travel.

2. The high-precision railway map generation method based on one-way trajectory correction according to claim 1 is characterized in that: After simultaneously collecting the continuous image sequence and the spatial posture data of the current train during its travel, the method further includes: Preprocessing the continuous image sequence; the preprocessing includes at least one of distortion correction, denoising and image enhancement; Filtering binocular images in the continuous image sequence based on timestamps of the left-eye image and the right-eye image in the continuous image sequence; The spatial posture data is synchronized and calibrated based on the timestamps of the binocular images in the continuous image sequence.

3. The high-precision railway map generation method based on one-way trajectory correction according to claim 1 is characterized in that: The synchronous positioning and map construction based on the posture change data and the spatial posture data to generate an initial electronic map includes: Determining a state vector based on actual measurement values ​​of each sensor; the sensor includes a binocular camera for collecting the continuous image sequence and an inertial measurement unit for collecting the spatial posture data; Determine a state estimation cost function based on actual measurement values ​​and expected calculated values ​​of each sensor at each moment, and a covariance matrix of the actual measurement values; Optimizing and solving the state estimation cost function to obtain an optimal estimation value of the state vector of the current train during its travel; Based on the optimal estimated value of the state vector, synchronous positioning and map construction are performed to generate an initial electronic map.

4. The high-precision railway map generation method based on one-way trajectory correction according to claim 1 is characterized in that: The method of correcting the mileage and track of the current train in the initial electronic map based on the mileage information and image positions corresponding to the preset markers in the continuous image sequence to generate an electronic map corresponding to the current train during its travel includes: Performing landmark recognition on the continuous image sequence to determine the target type of each preset landmark in the continuous image sequence; Based on the target type of each preset marker, determine the mileage information corresponding to each preset marker; The mileage information corresponding to each preset marker is compared with the measured mileage of the current train, and the mileage of the current train in the initial electronic map is corrected based on the mileage deviation between the mileage information and the measured mileage.

5. The high-precision railway map generation method based on one-way trajectory correction according to claim 1 is characterized in that: The method of correcting the mileage and track of the current train in the initial electronic map based on the mileage information and image positions corresponding to the preset markers in the continuous image sequence to generate an electronic map corresponding to the current train during its travel includes: Performing landmark recognition on the continuous image sequence to determine the image position of each preset landmark in the continuous image sequence; Based on the image position and the spatial coordinate position of each preset marker, and the parameters of the binocular camera that collects the continuous image sequence, determine the actual position information of the binocular camera; Based on the actual position information of the binocular camera, the track of the current train in the initial electronic map is corrected.

6. The high-precision railway map generation method based on one-way trajectory correction according to claim 4 or 5, characterized in that: The performing of landmark recognition on the continuous image sequence comprises: Inputting the continuous image sequence into a landmark recognition model to obtain the target type, image position and confidence of each preset landmark output by the landmark recognition model; The marker recognition model is obtained by training based on sample images corresponding to each preset marker.

7. The high-precision railway map generation method based on one-way trajectory correction according to claim 1 is characterized in that: After correcting the mileage and track of the current train in the initial electronic map based on the mileage information and image positions corresponding to the preset markers in the continuous image sequence to generate an electronic map corresponding to the current train during its travel, the method further includes: Determine the latitude and longitude coordinates of the starting point of the current train in the electronic map; Determine the longitude and latitude coordinates of each track point based on the longitude and latitude coordinates of the starting point and the relative displacement data of each track point of the current train; Based on the longitude and latitude coordinates, the track point number, the road section name, the road section type and the road section length of each track point, a key point table, a road section table and a road section details table corresponding to the electronic map are generated.

8. A high-precision railway map generation system based on one-way track correction, characterized in that: include: A data acquisition module is used to simultaneously acquire the continuous image sequence and spatial posture data of the current train during its travel; A feature extraction module, used for determining the posture change data of the current train based on the visual feature points in the continuous image sequence; A synchronous mapping module, used for performing synchronous positioning and mapping based on the posture change data and the spatial posture data to generate an initial electronic map; The map correction module is used to correct the mileage and track of the current train in the initial electronic map based on the mileage information and image positions corresponding to the preset markers in the continuous image sequence, and generate an electronic map corresponding to the current train during its travel.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method for generating a high-precision railway map based on one-way trajectory correction as described in any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for generating a high-precision railway map based on one-way track correction as described in any one of claims 1 to 7 is implemented.

Images