Method and system for quickly positioning railway invasion and limit personnel based on visual estimation

Abstract

The invention provides a method and a system for quickly positioning railway intrusion personnel based on visual estimation, belongs to the technical field of target identification, applies an advanced monocular depth estimation and positioning method to rail transit, and solves the potential safety hazard of train operation caused by insufficient personnel intrusion position along the current railway. According to a pre-constructed target prior mapping relation between image pixel coordinates and actual geographic coordinates, after an invader is detected, the pixel coordinates of the invader are directly mapped to corresponding actual longitude and latitude positions; and if the target prior mapping information does not exist, calling a coordinate conversion algorithm according to the pixel coordinates to realize conversion from the pixel coordinates to latitude and longitude coordinates, realizing dynamic conversion from an image plane to a global geographic coordinate system, and obtaining actual latitude and longitude positioning of personnel invasion. According to the invention, high-precision positioning can be carried out on the invading personnel in the rail transit scene only by means of the monocular camera, so that the problem that personnel invasion is difficult to check in an existing system is solved, and the safe operation level of a railway is improved.

Description

Technical Field

[0001] This invention relates to the field of target recognition technology, specifically to a method and system for rapid location of personnel encroaching on railway tracks based on visual estimation. Background Technology

[0002] To better ensure the efficiency and safety of railway transportation, the railway line environment needs real-time location tracking and alarm systems for personnel encroaching on railway lines, enabling timely protective measures to prevent accidents. Currently, real-world location tracking typically relies on the fusion of visual detection and lidar range measurement information to obtain the target's spatial location information.

[0003] However, the above positioning methods have significant shortcomings in railway line environments. On the one hand, railway line scenes are typically relatively simple and lack salient features. The reflection characteristics of lidar point clouds on tracks, ballast, and vegetation on both sides of the line are similar, leading to a decrease in spatial positioning accuracy. On the other hand, visual information is easily interfered with under complex lighting conditions such as strong light, rain, snow, fog, and nighttime, resulting in blurred or missing image features, making it difficult to guarantee the reliability of visual and lidar fusion. In addition, the long and narrow railway lines and the large changes in terrain result in high costs for sensor synchronization and calibration maintenance, further limiting the practical application effect in railway scenarios.

[0004] In vision-localization tasks, if camera intrinsic parameters (such as focal length, principal point, distortion, etc.) cannot be accurately obtained, or if the intrinsic parameters cannot be assumed to be known (i.e., "unknown intrinsic parameters" or "weak intrinsic parameters"), the localization system usually has to rely on depth or 3D structural information estimated from the image. Depth estimation (especially monocular) is inherently constrained by various factors such as texture, parallax, occlusion, lighting variations, and weather / nighttime conditions, making these methods often unstable. With the development of computer vision and deep learning in recent years, especially the introduction of Transformer architecture, multi-view learning, self-supervised depth estimation, and end-to-end reconstruction techniques, this instability is gradually improving. For example, BA-Net achieves joint depth and pose estimation, reducing the risk of error accumulation in traditional processes. Recent image intrinsic and extrinsic parameter estimation goes a step further, employing alternating intra-frame attention and global attention mechanisms to directly predict various geometric quantities such as camera intrinsic and extrinsic parameters, depth maps, point clouds, and point trajectories from one or more frames of input, achieving excellent performance even under conditions where intrinsic parameters are unknown. Although these methods have been theoretically validated, they still cannot fully meet the needs of field applications in engineering practice. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for rapid location of personnel encroaching on railway tracks based on visual estimation, so as to solve at least one of the technical problems existing in the background art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] In a first aspect, the present invention provides a method for rapid location of personnel encroaching on railway tracks based on visual estimation, comprising:

[0008] Acquire surveillance video streams from various sections along the rail transit line and extract corresponding basic video information, including video frame rate, resolution, camera number, and acquisition time.

[0009] The monitoring video stream is subjected to target detection analysis; when the target detection algorithm detects a person target in the current video frame, the identification information of the current video stream and the bounding box pixel coordinates of the person target are output.

[0010] The location information of the surveillance camera corresponding to the video stream and the preset prior mapping information are obtained, and the information is input into the coordinate transformation model. The coordinate transformation model directly calculates the latitude and longitude coordinates of the target in the actual geographic space based on the pixel coordinates of the personnel target.

[0011] If there is no preset prior mapping information, the image intrinsic and extrinsic parameter estimation model is called to perform feedforward estimation on the video image in order to predict the intrinsic and extrinsic parameter matrices of the camera.

[0012] It receives the camera intrinsic parameter matrix, extrinsic parameter matrix and distortion parameters estimated from the image intrinsic and extrinsic parameter estimation model, adaptively corrects the camera model parameters, and updates and fits each parameter in real time based on the parameter set;

[0013] Based on the updated camera extrinsic parameters, a geometric mapping method is used to project the two-dimensional image coordinate points into the three-dimensional camera coordinate space to obtain the three-dimensional position coordinates of the person target in the camera space, thus achieving accurate mapping from the image plane to the three-dimensional space.

[0014] The camera spatial coordinates of personnel targets are mapped to geographic latitude and longitude coordinates in real time, thereby realizing the dynamic transformation from the image plane to the global geographic coordinate system and obtaining the actual latitude and longitude positioning of personnel intrusion.

[0015] As a further limitation of the first aspect of the present invention, the training process of the image intrinsic and extrinsic parameter estimation model is as follows: a multi-view training set of rail transit containing real annotations is constructed, and each training sample consists of the following three parts: a multi-view image set, camera intrinsic parameter labels, and camera extrinsic parameter labels; the image encoder encodes each image into a region-level feature; the features of multiple views are modeled in a cross-view correlation to capture geometric consistency; the camera parameter prediction head predicts the intrinsic and extrinsic parameters; the focal length and principal point are predicted; and the pose is predicted.

[0016] As a further limitation of the first aspect of the invention, the camera extrinsic parameter labels use rotation and translation in the world coordinate system.

[0017] As a further limitation of the first aspect of the present invention, the loss function for the training process of the image intrinsic and extrinsic parameter estimation model is designed as follows: the total loss function is the sum of intrinsic parameter regression loss, extrinsic parameter regression loss, geometric reprojection loss, and reduction of reprojection error.

[0018] As a further limitation of the first aspect of the present invention, the optimization algorithm focuses on minimizing the reprojection error and constructs the objective function based on the projection points calculated from the front camera parameters.

[0019] As a further limitation of the first aspect of the present invention, the optimization adopts an iterative method based on Jacobi. By linearizing the projection model, calculating the residual and its Jacobi matrix, the camera parameters are continuously updated so that the projection error gradually converges. Among them, the robust loss function suppresses outliers so that the optimization process remains stable under noise interference or local anomalies, and finally the converged optimal intrinsic parameters, extrinsic parameters and distortion parameters are obtained.

[0020] Secondly, the present invention provides a rapid positioning system for railway personnel encroaching on railway lines based on visual estimation, comprising:

[0021] The acquisition module is used to acquire the surveillance video streams of each section along the rail transit line and extract the corresponding basic video information, including video frame rate, resolution, camera number and acquisition time.

[0022] The detection module is used to perform target detection analysis on the monitoring video stream; when the target detection algorithm detects a person target in the current video frame, it outputs the identification information of the current video stream and the bounding box pixel coordinates of the person target;

[0023] The conversion module is used to obtain the location information of the surveillance camera corresponding to the video stream and the preset prior mapping information, and input the information into the coordinate conversion model. The coordinate conversion model directly calculates the latitude and longitude coordinates of the target in the actual geographic space based on the pixel coordinates of the personnel target.

[0024] The estimation module is used to perform feedforward estimation of the video image by calling the image intrinsic and extrinsic parameter estimation model if there is no preset prior mapping information, so as to predict the intrinsic and extrinsic parameter matrices of the camera.

[0025] The correction module is used to receive the camera intrinsic parameter matrix, extrinsic parameter matrix and distortion parameters estimated by the image intrinsic and extrinsic parameter estimation model, adaptively correct the camera model parameters, and update and fit each parameter in real time based on the parameter set.

[0026] The projection module is used to project two-dimensional image coordinate points into the three-dimensional camera coordinate space using a geometric mapping method based on the updated camera extrinsic parameters, thereby obtaining the three-dimensional position coordinates of the person target in the camera space and realizing accurate mapping from the image plane to the three-dimensional space.

[0027] The positioning module is used to map the camera spatial coordinates of personnel targets to geographic latitude and longitude coordinates in real time, thereby realizing the dynamic transformation from the image plane to the global geographic coordinate system and obtaining the actual latitude and longitude positioning of personnel intrusion.

[0028] Secondly, the present invention provides a rapid positioning system for railway personnel encroaching on railway lines based on visual estimation, comprising:

[0029] The acquisition module is used to acquire the surveillance video streams of each section along the rail transit line and extract the corresponding basic video information, including video frame rate, resolution, camera number and acquisition time.

[0030] The detection module is used to perform target detection analysis on the monitoring video stream; when the target detection algorithm detects a person target in the current video frame, it outputs the identification information of the current video stream and the bounding box pixel coordinates of the person target;

[0031] The conversion module is used to obtain the location information of the surveillance camera corresponding to the video stream and the preset prior mapping information, and input the information into the coordinate conversion model. The coordinate conversion model directly calculates the latitude and longitude coordinates of the target in the actual geographic space based on the pixel coordinates of the personnel target.

[0032] The estimation module is used to perform feedforward estimation of the video image by calling the image intrinsic and extrinsic parameter estimation model if there is no preset prior mapping information, so as to predict the intrinsic and extrinsic parameter matrices of the camera.

[0033] The correction module is used to receive the camera intrinsic parameter matrix, extrinsic parameter matrix and distortion parameters estimated by the image intrinsic and extrinsic parameter estimation model, adaptively correct the camera model parameters, and update and fit each parameter in real time based on the parameter set.

[0034] The projection module is used to project two-dimensional image coordinate points into the three-dimensional camera coordinate space using a geometric mapping method based on the updated camera extrinsic parameters, thereby obtaining the three-dimensional position coordinates of the person target in the camera space and realizing accurate mapping from the image plane to the three-dimensional space.

[0035] The positioning module is used to map the camera spatial coordinates of personnel targets to geographic latitude and longitude coordinates in real time, thereby realizing the dynamic transformation from the image plane to the global geographic coordinate system and obtaining the actual latitude and longitude positioning of personnel intrusion.

[0036] Thirdly, the present invention provides a non-transitory computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the method for rapid location of railway encroaching personnel based on visual estimation as described in the first aspect.

[0037] Fourthly, the present invention provides a computer device including a memory and a processor, wherein the processor and the memory communicate with each other, the memory stores program instructions that can be executed by the processor, and the processor invokes the program instructions to execute the method for rapid location of railway encroachment personnel based on visual estimation as described in the first aspect.

[0038] Fifthly, the present invention provides an electronic device comprising: a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing the method for rapid location of railway encroachment personnel based on visual estimation as described in the first aspect.

[0039] The beneficial effects of this invention are: it can accurately locate intruders in rail transit scenarios using only a monocular camera, thereby solving the problem of difficulty in investigating intrusions in existing systems and improving the level of railway safety operation.

[0040] The advantages of additional aspects of the invention will be set forth more clearly in the following description or will be learned by practice of the invention. Attached Figure Description

[0041] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a flowchart of the method for rapid location of railway personnel encroaching on railway tracks based on visual estimation, as described in an embodiment of the present invention.

[0043] Figure 2 This is a flowchart illustrating the actual corresponding points and model conversion points as described in an embodiment of the present invention. Detailed Implementation

[0044] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0045] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0046] It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with their meanings in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless defined as here.

[0047] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, and / or groups thereof.

[0048] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0049] To facilitate understanding of the present invention, the present invention will be further explained and described below with reference to the accompanying drawings and specific embodiments. However, the specific embodiments do not constitute a limitation on the embodiments of the present invention.

[0050] Those skilled in the art should understand that the accompanying drawings are merely schematic diagrams of embodiments, and the components in the drawings are not necessarily essential for implementing the present invention.

[0051] Example 1

[0052] In this embodiment 1, a rapid positioning system for railway personnel encroaching on railway tracks based on visual estimation is provided, comprising: an acquisition module for acquiring monitoring video streams from various sections along the railway line and extracting corresponding basic video information, including video frame rate, resolution, camera number, and acquisition time; a detection module for performing target detection analysis on the monitoring video stream; when the target detection algorithm detects a personnel target in the current video frame, it outputs the identification information of the current video stream and the bounding box pixel coordinates of the personnel target; a conversion module for acquiring the location information of the monitoring camera corresponding to the video stream and preset prior mapping information, and inputting the information into a coordinate conversion model, wherein the coordinate conversion model directly calculates the latitude and longitude coordinates of the target in the actual geographic space based on the pixel coordinates of the personnel target; and an estimation module for estimating the target if there is no preset prior mapping information. The system receives the camera's intrinsic and extrinsic parameters. It then calls an image intrinsic and extrinsic parameter estimation model to perform feedforward estimation of the video image, predicting the camera's intrinsic and extrinsic parameter matrices. A correction module receives the camera's intrinsic and extrinsic parameter matrices and distortion parameters estimated by the image intrinsic and extrinsic parameter estimation model, adaptively corrects the camera model parameters, and updates and fits each parameter in real time based on the parameter set. A projection module projects the two-dimensional image coordinates onto the three-dimensional camera coordinate space using geometric mapping based on the updated camera extrinsic parameters, obtaining the three-dimensional position coordinates of the person target in the camera space, achieving accurate mapping from the image plane to three-dimensional space. A positioning module maps the camera space coordinates of the person target to geographic latitude and longitude coordinates in real time, thereby achieving dynamic transformation from the image plane to the global geographic coordinate system, obtaining the actual latitude and longitude positioning of the person's intrusion.

[0053] In this embodiment, a method for rapid location of railway intruders based on visual estimation is implemented based on the above system, addressing the safety risks to train operation caused by insufficient personnel intrusion location along current railway lines. For example... Figure 1 , Figure 2 As shown, the method includes the following steps:

[0054] Step 1: Obtain the surveillance video streams of each section along the rail transit line and extract the corresponding basic video information, including video frame rate, resolution, camera number and acquisition time, etc.

[0055] Step 2: Perform target detection analysis on the surveillance video stream. The target detection algorithms include the YOLO series and R-CNN series algorithms based on the Convolutional Neural Network (CNN) architecture, and the DETR series and DINO algorithms based on the Transformer architecture. When the target detection algorithm detects a person target in the current video frame, it outputs the identification information of the current video stream and the bounding box pixel coordinates of the person target;

[0056] Step 3: Obtain the location information of the surveillance camera corresponding to the video stream and the preset prior mapping information, and input the information into the coordinate transformation model. The coordinate transformation model directly calculates the latitude and longitude coordinates of the target in the actual geographic space based on the pixel coordinates of the personnel target.

[0057] Step 4: If there is no preset prior mapping information, the image intrinsic and extrinsic parameter estimation model is called to perform feedforward estimation on the video image in order to predict the intrinsic and extrinsic parameter matrices of the camera.

[0058] Step 5: Receive the camera intrinsic parameter matrix, extrinsic parameter matrix and distortion parameters estimated from the image intrinsic and extrinsic parameter estimation model. Adaptively correct the camera model parameters through statistical sampling, optimization algorithms or sliding window fitting to compensate for the impact of environmental changes, camera shake or temperature drift on the accuracy of the camera geometric model. Update and fit each parameter in real time based on the parameter set.

[0059] Step 6: Based on the updated camera extrinsic parameters, use geometric mapping methods such as spatial affine transformation or projective transformation to project the two-dimensional image coordinate points into the three-dimensional camera coordinate space to obtain the three-dimensional position coordinates of the person target in the camera space, thus achieving accurate mapping from the image plane to the three-dimensional space.

[0060] Step 7: Based on the information obtained in Steps 5 and 6, the camera spatial coordinates of the personnel target are mapped to geographic latitude and longitude coordinates in real time, thereby realizing the dynamic transformation from the image plane to the global geographic coordinate system and obtaining the actual latitude and longitude positioning of the personnel intrusion.

[0061] In step 4, the training process of the image intrinsic and extrinsic parameter estimation model is as follows:

[0062] (1) Dataset acquisition and label composition

[0063] To train an image intrinsic and extrinsic parameter estimation model for railway monitoring scenarios, it is first necessary to construct a multi-view training set of rail transit images containing real-world annotations. Each training sample consists of the following three parts: the multi-view image set denoted as... Where N represents the number of visible images in the scene, and I represents an image. Camera intrinsic parameter labels: the camera intrinsic parameter matrix corresponding to each image:

[0064]

[0065] Tags include focal length , Main point , .

[0066] Camera extrinsic parameter labels (pose): using rotation and translation in world coordinates:

[0067] , ∈SO(3), ∈R 3 ;in, Represents the rotation matrix of the i-th camera relative to the world coordinate system; SO(3) represents the translation vector of the i-th camera relative to the world coordinate system; SO(3) represents the three-dimensional special orthogonal group, i.e., all groups satisfying: RTR=I, det(R)=1; R 3 This represents a three-dimensional Euclidean space, which is a vector space consisting of three real numbers.

[0068] (2) Training data preprocessing

[0069] Image normalization and resizing: Resize all images to a fixed resolution (384*384) and normalize them. = ;

[0070] Diverse data augmentations: lighting enhancement (brightness / contrast changes), random noise (adding common noises in rail transit scenarios), small random rotation / scaling (ensuring synchronized label updates or lossless operation), blurring, and rain / snow texture enhancement (improving the robustness of actual deployment).

[0071] View collection organization: The model requires a collection of views as input, therefore, it needs to be organized for each scene:

[0072] ;

[0073] (3) Network structure

[0074] The image intrinsic and extrinsic parameter estimation model is a geometric learning model with Transformer as its core, consisting of the following modules:

[0075] Image encoder (Vision Transformer): Encodes each image into patch-level features.

[0076] ;

[0077] Multi-view geometry Transformer: Models cross-view correlations of features from multiple views, capturing geometric consistency: F = G - Former(F1, F2, ... F N ); G-Former(*) represents the geometry-aware Transformer module.

[0078] Camera Head: Directly predicts intrinsic and extrinsic parameters.

[0079] Predicted focal length and principal point: ;

[0080] Predicted pose (rotation and translation): ( )=g pose (F i );

[0081] (4) Loss function design

[0082] Internal reference regression loss: ;

[0083] External parameter regression loss: ;

[0084] Where dR represents the rotational difference.

[0085] Geometric reprojection loss (core):

[0086] For 3D point X provided by COLMAP / MVS j Reprojected onto the image plane:

[0087]

[0088]

[0089] Reduce reprojection error:

[0090] Final loss: .

[0091] In step 5, the optimization algorithm focuses on minimizing the reprojection error by constructing an objective function. ,in u represents the actual pixel. These are the projection points calculated based on the current camera parameters. Optimization employs an iterative approach based on Jacobi, which involves linearizing the projection model and calculating the residuals. and its Jacobian matrix J, then according to Continuously update camera parameters to gradually converge the projection error. Robust loss function. Used to suppress outliers, ensuring the optimization process remains stable under noise interference or local anomalies, ultimately yielding the converged optimal intrinsic parameters, extrinsic parameters, and distortion parameters.

[0092] The proposed method for rapid location of intruders in railway lines based on visual estimation enables high-precision location of intruders in rail transit scenarios using only a monocular camera, thereby solving the problem of difficulty in investigating intrusions in existing systems and improving the level of railway safety operation.

[0093] Specifically, in this embodiment, two methods are designed for different field application conditions: Method 1 is a PnP mapping positioning method based on preset point information and Method 2 is a latitude and longitude positioning method based on model feedforward estimation.

[0094] Among them, the PnP mapping localization method based on preset point information is suitable for situations where the installation location of the surveillance camera and the geometric information of the scene are known. First, the video stream from the rail transit surveillance camera is used as input and fed into the YOLOv12 detection algorithm to detect intruders and output the pixel coordinates of the selected person. Second, the preset point information shown in Table 1 (including camera intrinsic and extrinsic parameters and spatial reference point coordinates) is loaded. Finally, a coordinate transformation algorithm is called to convert the pixel coordinates into corresponding actual latitude and longitude coordinates according to the PnP mapping relationship, thus achieving person localization.

[0095] surface Preset location information

[0096]

[0097] The latitude and longitude localization method based on model feedforward estimation is suitable for application environments without preset camera parameters or scene information. First, the surveillance video stream is input into the YOLOv12 detection algorithm to detect and output the pixel coordinates of the person selected. Second, the video stream is simultaneously input into an image intrinsic and extrinsic parameter estimation model, which uses its pre-trained network to estimate camera intrinsic and extrinsic parameters and depth map values ​​from the video data. Finally, the estimated parameter matrix and the approximate latitude and longitude information of the surveillance camera are input into a coordinate transformation algorithm to output the actual latitude and longitude coordinates of the intruder.

[0098] In this embodiment, five intruders from the rail transit monitoring screen were selected as test subjects for the above method. Their detected pixel coordinates and actual latitude and longitude coordinates (obtained by GNSS measurement) were recorded. Positioning calculations were performed using PnP mapping positioning based on preset point information and latitude and longitude positioning based on model feedforward estimation, respectively. The results are shown in Table 2.

[0099] surface Location coordinate comparison

[0100]

[0101] Experimental results show that the visual estimation-based rapid positioning method for railway intrusion personnel proposed in this embodiment can achieve high-precision positioning under different conditions. The two embodiments demonstrate that the two types of methods in this invention can be flexibly switched in the system, ensuring high precision while possessing good versatility and scalability, providing an efficient, low-cost, and engineering-oriented solution for railway intrusion detection.

[0102] Example 3

[0103] This embodiment 3 provides a non-transitory computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, they implement the visual estimation-based rapid location method for railway encroachment personnel as described above. The method includes:

[0104] Acquire surveillance video streams from various sections along the rail transit line and extract corresponding basic video information, including video frame rate, resolution, camera number, and acquisition time.

[0105] The monitoring video stream is subjected to target detection analysis; when the target detection algorithm detects a person target in the current video frame, the identification information of the current video stream and the bounding box pixel coordinates of the person target are output.

[0106] The location information of the surveillance camera corresponding to the video stream and the preset prior mapping information are obtained, and the information is input into the coordinate transformation model. The coordinate transformation model directly calculates the latitude and longitude coordinates of the target in the actual geographic space based on the pixel coordinates of the personnel target.

[0107] If there is no preset prior mapping information, the image intrinsic and extrinsic parameter estimation model is called to perform feedforward estimation on the video image in order to predict the intrinsic and extrinsic parameter matrices of the camera.

[0108] It receives the camera intrinsic parameter matrix, extrinsic parameter matrix and distortion parameters estimated from the image intrinsic and extrinsic parameter estimation model, adaptively corrects the camera model parameters, and updates and fits each parameter in real time based on the parameter set;

[0109] Based on the updated camera extrinsic parameters, a geometric mapping method is used to project the two-dimensional image coordinate points into the three-dimensional camera coordinate space to obtain the three-dimensional position coordinates of the person target in the camera space, thus achieving accurate mapping from the image plane to the three-dimensional space.

[0110] The camera spatial coordinates of personnel targets are mapped to geographic latitude and longitude coordinates in real time, thereby realizing the dynamic transformation from the image plane to the global geographic coordinate system and obtaining the actual latitude and longitude positioning of personnel intrusion.

[0111] Example 4

[0112] This embodiment 4 provides a computer device, including a memory and a processor, wherein the processor and the memory communicate with each other, and the memory stores program instructions that can be executed by the processor. The processor calls the program instructions to execute the visual estimation-based rapid location method for railway encroachment personnel as described above, the method including:

[0113] Acquire surveillance video streams from various sections along the rail transit line and extract corresponding basic video information, including video frame rate, resolution, camera number, and acquisition time.

[0114] The monitoring video stream is subjected to target detection analysis; when the target detection algorithm detects a person target in the current video frame, the identification information of the current video stream and the bounding box pixel coordinates of the person target are output.

[0115] The location information of the surveillance camera corresponding to the video stream and the preset prior mapping information are obtained, and the information is input into the coordinate transformation model. The coordinate transformation model directly calculates the latitude and longitude coordinates of the target in the actual geographic space based on the pixel coordinates of the personnel target.

[0116] If there is no preset prior mapping information, the image intrinsic and extrinsic parameter estimation model is called to perform feedforward estimation on the video image in order to predict the intrinsic and extrinsic parameter matrices of the camera.

[0117] It receives the camera intrinsic parameter matrix, extrinsic parameter matrix and distortion parameters estimated from the image intrinsic and extrinsic parameter estimation model, adaptively corrects the camera model parameters, and updates and fits each parameter in real time based on the parameter set;

[0118] Based on the updated camera extrinsic parameters, a geometric mapping method is used to project the two-dimensional image coordinate points into the three-dimensional camera coordinate space to obtain the three-dimensional position coordinates of the person target in the camera space, thus achieving accurate mapping from the image plane to the three-dimensional space.

[0119] The camera spatial coordinates of personnel targets are mapped to geographic latitude and longitude coordinates in real time, thereby realizing the dynamic transformation from the image plane to the global geographic coordinate system and obtaining the actual latitude and longitude positioning of personnel intrusion.

[0120] Example 5

[0121] This embodiment 5 provides an electronic device, including: a processor, a memory, and a computer program; wherein, the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing the visual estimation-based rapid location method for railway encroachment personnel as described above. The method includes:

[0122] Acquire surveillance video streams from various sections along the rail transit line and extract corresponding basic video information, including video frame rate, resolution, camera number, and acquisition time.

[0123] The monitoring video stream is subjected to target detection analysis; when the target detection algorithm detects a person target in the current video frame, the identification information of the current video stream and the bounding box pixel coordinates of the person target are output.

[0124] The location information of the surveillance camera corresponding to the video stream and the preset prior mapping information are obtained, and the information is input into the coordinate transformation model. The coordinate transformation model directly calculates the latitude and longitude coordinates of the target in the actual geographic space based on the pixel coordinates of the personnel target.

[0125] If there is no preset prior mapping information, the image intrinsic and extrinsic parameter estimation model is called to perform feedforward estimation on the video image in order to predict the intrinsic and extrinsic parameter matrices of the camera.

[0126] It receives the camera intrinsic parameter matrix, extrinsic parameter matrix and distortion parameters estimated from the image intrinsic and extrinsic parameter estimation model, adaptively corrects the camera model parameters, and updates and fits each parameter in real time based on the parameter set;

[0127] Based on the updated camera extrinsic parameters, a geometric mapping method is used to project the two-dimensional image coordinate points into the three-dimensional camera coordinate space to obtain the three-dimensional position coordinates of the person target in the camera space, thus achieving accurate mapping from the image plane to the three-dimensional space.

[0128] The camera spatial coordinates of personnel targets are mapped to geographic latitude and longitude coordinates in real time, thereby realizing the dynamic transformation from the image plane to the global geographic coordinate system and obtaining the actual latitude and longitude positioning of personnel intrusion.

[0129] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0130] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0131] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0132] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment, whereby a series of operational steps are performed to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0133] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that, based on the technical solutions disclosed in the present invention, various modifications or variations that can be made by those skilled in the art without creative effort should be included within the scope of protection of the present invention.

Claims

1. A method for rapid location of personnel encroaching on railway tracks based on visual estimation, characterized in that, include: Acquire surveillance video streams from various sections along the rail transit line and extract corresponding basic video information, including video frame rate, resolution, camera number, and acquisition time. The surveillance video stream is subjected to target detection and analysis processing; When the target detection algorithm detects a person in the current video frame, it outputs the identification information of the current video stream and the bounding box pixel coordinates of the person. The location information of the surveillance camera corresponding to the video stream and the preset prior mapping information are obtained, and the information is input into the coordinate transformation model. The coordinate transformation model directly calculates the latitude and longitude coordinates of the target in the actual geographic space based on the pixel coordinates of the personnel target. If there is no preset prior mapping information, the image intrinsic and extrinsic parameter estimation model is called to perform feedforward estimation on the video image in order to predict the intrinsic and extrinsic parameter matrices of the camera. It receives the camera intrinsic parameter matrix, extrinsic parameter matrix and distortion parameters estimated from the image intrinsic and extrinsic parameter estimation model, adaptively corrects the camera model parameters, and updates and fits each parameter in real time based on the parameter set; Based on the updated camera extrinsic parameters, a geometric mapping method is used to project the two-dimensional image coordinate points into the three-dimensional camera coordinate space to obtain the three-dimensional position coordinates of the person target in the camera space, thus achieving accurate mapping from the image plane to the three-dimensional space. The camera spatial coordinates of personnel targets are mapped to geographic latitude and longitude coordinates in real time, thereby realizing the dynamic transformation from the image plane to the global geographic coordinate system and obtaining the actual latitude and longitude positioning of personnel intrusion.

2. The method for rapid location of personnel encroaching on railway tracks based on visual estimation according to claim 1, characterized in that, The training process of the image intrinsic and extrinsic parameter estimation model is as follows: Construct a multi-view training set of rail transit with real annotations. Each training sample consists of the following three parts: a set of multi-view images, camera intrinsic parameter labels, and camera extrinsic parameter labels; The image encoder encodes each image into region-level features; The features of multiple views are modeled in a cross-view correlation to capture geometric consistency; The camera parameter prediction head predicts intrinsic and extrinsic parameters; Predicts focal length and principal point; Predicts pose.

3. The method for rapid location of personnel encroaching on railway tracks based on visual estimation according to claim 2, characterized in that, The camera extrinsic labels use rotation and translation in the world coordinate system.

4. The method for rapid location of railway personnel encroaching on railway tracks based on visual estimation according to claim 2, characterized in that, The loss function for training the image intrinsic and extrinsic parameter estimation model is designed as follows: the total loss function is the sum of intrinsic parameter regression loss, extrinsic parameter regression loss, geometric reprojection loss, and loss to reduce reprojection error.

5. The method for rapid location of personnel encroaching on railway tracks based on visual estimation according to claim 1, characterized in that, The optimization algorithm focuses on minimizing the reprojection error and constructs the objective function based on the projection points calculated from the front camera parameters.

6. The method for rapid location of personnel encroaching on railway tracks based on visual estimation according to claim 5, characterized in that, The optimization adopts an iterative approach based on Jacobi. By linearizing the projection model, calculating the residuals and their Jacobi matrices, the camera parameters are continuously updated, and the projection error gradually converges. In particular, a robust loss function is used to suppress outliers, so that the optimization process remains stable under noise interference or local anomalies, and finally the converged optimal intrinsic parameters, extrinsic parameters and distortion parameters are obtained.

7. A rapid positioning system for personnel encroaching on railway tracks based on visual estimation, characterized in that, include: The acquisition module is used to acquire surveillance video streams from various sections along the rail transit line and extract the corresponding basic video information, including video frame rate, resolution, camera number, and acquisition time. The detection module is used to perform target detection and analysis processing on the monitoring video stream; When the target detection algorithm detects a person in the current video frame, it outputs the identification information of the current video stream and the bounding box pixel coordinates of the person. The conversion module is used to obtain the location information of the surveillance camera corresponding to the video stream and the preset prior mapping information, and input the information into the coordinate conversion model. The coordinate conversion model directly calculates the latitude and longitude coordinates of the target in the actual geographic space based on the pixel coordinates of the personnel target. The estimation module is used to perform feedforward estimation of the video image by calling the image intrinsic and extrinsic parameter estimation model if there is no preset prior mapping information, so as to predict the intrinsic and extrinsic parameter matrices of the camera. The correction module is used to receive the camera intrinsic parameter matrix, extrinsic parameter matrix and distortion parameters estimated from the image intrinsic and extrinsic parameter estimation model, adaptively correct the camera model parameters, and update and fit each parameter in real time based on the parameter set. The projection module is used to project two-dimensional image coordinate points into the three-dimensional camera coordinate space using a geometric mapping method based on the updated camera extrinsic parameters, thereby obtaining the three-dimensional position coordinates of the person target in the camera space and realizing accurate mapping from the image plane to the three-dimensional space. The positioning module is used to map the camera spatial coordinates of personnel targets to geographic latitude and longitude coordinates in real time, thereby realizing the dynamic transformation from the image plane to the global geographic coordinate system and obtaining the actual latitude and longitude positioning of personnel intrusion.

8. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium is used to store computer instructions, which, when executed by a processor, implement the method for rapid location of railway encroaching personnel based on visual estimation as described in any one of claims 1-6.

9. A computer device, characterized in that, The system includes a memory and a processor, which communicate with each other. The memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the method for rapid location of railway encroaching personnel based on visual estimation as described in any one of claims 1-6.

10. An electronic device, characterized in that, include: The device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing the method for rapid location of railway encroaching personnel based on visual estimation as described in any one of claims 1-6.

Images