A method for identifying foreign matter intrusion into a subway line

Abstract

The application discloses a metro line foreign matter intrusion identification method, which comprises the following steps: setting a trackside monitoring, positioning, calculating and communicating device along a metro track line, and setting an image detection device and a radar detection device on a metro car head for data collection; then, real-time visual field image in the collected data is identified and detected to obtain a visual field image feature detection result; real-time radar data is identified and monitored to obtain a current radar detection object position; according to a data synchronization frequency, the current radar detection object position after coordinate conversion and the visual field image feature detection result are fused in target position to obtain a vehicle-mounted intrusion detection result; the vehicle-mounted intrusion detection result is uploaded by using a metro vehicle-mounted communication device, combined with a trackside monitoring foreign matter intrusion result, and mapping alarm is performed on a metro track line simulation map.

Description

A method for identifying foreign object intrusion in subway lines Technical Field

[0001] This invention belongs to the field of subway track safety technology, and more specifically, relates to a method for identifying foreign object intrusion into subway lines. Background Technology

[0002] Encroachment specifically refers to the situation where an object exceeds the permitted safe dimensions within the safety limits of rail transit operations, thereby interfering with the normal operation of trains on that section of the line. In recent years, urban rail transit accidents caused by object encroachment have become increasingly common.

[0003] Urban rail transit lines are typically laid within tunnels, and due to route selection constraints, they often feature small-radius curves and steep gradients with significant gradient variations and complex alignments, posing numerous challenges to the development of autonomous obstacle detection systems for trains. Existing urban rail transit lines primarily rely on millimeter-wave radar, lidar, or visual and acoustic sensors for obstacle intrusion detection. While this reduces the likelihood of missed obstacles, its effective detection range is relatively short, and blind spots exist, resulting in insufficient accuracy and efficiency in obstacle identification and localization. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for identifying foreign object encroachment in subway lines, which solves the problem that the existing detection of foreign object encroachment in urban rail transit is not comprehensive and accurate enough.

[0005] To achieve the above-mentioned objective, the present invention provides a method for identifying foreign object intrusion in subway lines, characterized by comprising the following steps:

[0006] (1) Set up trackside monitoring, positioning, calculation and communication devices along the subway track line to identify and upload the results of foreign object intrusion monitoring along the trackside and to construct a simulated map of the subway track line.

[0007] (2) The image detection device and the radar detection device are set at the front of the subway car to capture real-time visual images of the subway's direction of travel and to detect real-time radar data of the subway's direction of travel.

[0008] (3) Use the visual field image foreign object detection neural network in the subway vehicle computing device to identify and detect real-time visual field images, and obtain the visual field image feature detection results;

[0009] (4) Based on the Kalman filter algorithm, the subway onboard computing device is used to identify and monitor the real-time radar data to obtain the current radar detection object position.

[0010] (5) Based on the data synchronization frequency, the target position is fused by the current radar-detected object position and the field-of-view image feature detection results using the subway vehicle-mounted computing device to obtain the vehicle-mounted intrusion detection results.

[0011] (6) The subway onboard computing device combines the onboard intrusion detection results and the trackside monitoring foreign object intrusion results to map and alarm on the subway track line simulation map.

[0012] The objective of this invention is achieved as follows:

[0013] This invention discloses a method for identifying foreign object intrusions on subway lines. The method involves installing trackside monitoring, positioning, calculation, and communication devices along the subway track, and image detection and radar detection devices at the front of the subway car for data acquisition. The method then identifies and detects real-time field-of-view images from the acquired data to obtain field-of-view image feature detection results. Real-time radar data is also monitored to determine the current radar-detected object position. Based on the data synchronization frequency, the coordinate-transformed current radar-detected object position and the field-of-view image feature detection results are fused to obtain the vehicle-mounted intrusion detection result. The vehicle-mounted intrusion detection result is uploaded using the subway vehicle-mounted communication device and, combined with the trackside monitoring foreign object intrusion result, is mapped and alarmed on a simulated subway track map.

[0014] Meanwhile, the method for identifying foreign object intrusion in subway lines according to the present invention also has the following beneficial effects:

[0015] (1) By setting up trackside monitoring, positioning, calculation and communication devices on the subway track, as well as image detection devices and radar detection devices at the front of the subway car, the present invention realizes all-round and multi-dimensional monitoring of foreign objects in the direction of subway travel;

[0016] (2) This invention combines trackside monitoring with vehicle-mounted monitoring, which improves the accuracy and reliability of foreign object detection and effectively reduces safety accidents caused by foreign object intrusion.

[0017] (3) This invention utilizes a field-of-view image foreign object detection neural network to identify and detect real-time field-of-view images, a Kalman filter algorithm to process radar data, and, based on the data synchronization frequency, fuses the radar-detected object position after coordinate transformation with the field-of-view image feature detection results to improve the accuracy and precision of foreign object detection.

[0018] (4) This invention constructs a simulated map of subway track lines and maps alarm information on the map, realizing the visual management of foreign object intrusion. It is convenient to intuitively understand the situation of foreign object intrusion, make quick decisions and take corresponding countermeasures, and effectively improve the speed of emergency response and management efficiency. Attached Figure Description

[0019] Figure 1 is a flowchart of a method for identifying foreign object intrusion in subway lines according to the present invention;

[0020] Figure 2 is a block diagram of a foreign object intrusion monitoring system for subway lines according to the present invention. Detailed Implementation

[0021] The specific embodiments of the present invention will now be described with reference to the accompanying drawings to enable those skilled in the art to better understand the invention. It should be particularly noted that in the following description, detailed descriptions of known functions and designs that might obscure the main content of the invention will be omitted here.

[0022] Example

[0023] Figure 1 is a flowchart of a method for identifying foreign object intrusion in subway lines according to the present invention.

[0024] In this embodiment, as shown in Figure 1, the method for identifying foreign object intrusion in subway lines according to the present invention includes the following steps:

[0025] (1) Set up trackside monitoring, positioning, calculation and communication devices along the subway track line to identify and upload the results of foreign object intrusion monitoring along the trackside and to construct a simulated map of the subway track line.

[0026] (1.1) According to the preset first interval distance, trackside monitoring devices and trackside computing devices connected to trackside monitoring devices are installed sequentially along the platforms, curves and connecting passages of the subway track line. The trackside monitoring devices are used to capture trackside monitoring images; the trackside computing devices are used to identify the foreign object intrusion results in the trackside monitoring images.

[0027] In this embodiment, the trackside monitoring device uses a long-focal-length laser high-definition camera. The first interval distance allows the trackside monitoring device to cover areas prone to intrusion, such as platforms, curves, and connecting passages, without blind spots. Considering the intrusion situations of foreign objects of different sizes, to avoid false detection due to magnification of larger objects and missed detection of smaller objects, enhanced target detection at different scales is performed on the trackside monitoring images. This enables the trackside computing device to quickly and accurately identify the intrusion results in the trackside monitoring images. The specific method is as follows:

[0028] (1.1.1) Acquire several trackside monitoring images to form a trackside monitoring image dataset;

[0029] (1.1.2) Keep the backbone and neck network of the YOLOv5 network unchanged, replace the three prediction head sub-networks in the prediction head network with the classification prediction sub-network, the size prediction sub-network and the offset prediction sub-network to obtain the trackside image intrusion recognition network, and then set the trackside image intrusion recognition network in the trackside computing device.

[0030] In this embodiment, the prediction head network uses three parallel convolutional blocks as classification prediction sub-networks, size prediction sub-networks, and offset prediction sub-networks, respectively, outputting three feature maps with the same resolution. The feature map output by the classification prediction sub-network can predict the category and location information of the intruding target. Different channels represent different categories, and the matrix element values ​​in the feature map matrix are the confidence scores for points centered on the intruding target. The feature map output by the size prediction sub-network has two channels, corresponding to the width and height of the intruding target. The feature map output by the offset prediction sub-network also has two channels, corresponding to the horizontal and vertical offsets of the center pixel of the intruding target.

[0031] (1.1.3) Based on the multi-scale target loss function, the trackside monitoring image is input into the trackside image intrusion recognition network for intrusion target recognition prediction training, and the trained trackside image intrusion recognition network is obtained.

[0032] In this embodiment, the multi-scale target loss function used by the trackside image intrusion recognition network during the training process is:

[0033] L = L1 + μ1L2 + μ2L3

[0034]

[0035] Where L represents the multi-scale target loss function, L1 represents the classification loss function, μ1 represents the first loss function hyperparameter, L2 represents the offset loss function, μ2 represents the second loss function hyperparameter, L3 represents the size loss function, N represents the total number of intruding targets, C represents the number of channels, H represents the height of the trackside monitoring image, W represents the width of the trackside monitoring image, c represents the class of pixels in the image, y represents the ordinate of a pixel in the image, and x represents the abscissa of a pixel in the image. υ represents the probability that a pixel belongs to the center point of a certain category. cyx This represents the true probability that a pixel belongs to the center point of a certain category, where α represents the first category exponent factor and β represents the second category exponent factor. θ represents the offset between the center point of the i-th intrusion target in the trackside monitoring image and the center point of the actual target bounding box in the trackside monitoring image. i Let ||·||1 represent the true offset between the center point of the i-th intrusion target in the trackside monitoring image and the center point of the actual target bounding box in the trackside monitoring image, and ||·||1 represent the first norm. ξ represents the size of the i-th intrusion target corresponding to the predicted center point. i This represents the actual size of the i-th intrusion target.

[0036] (1.1.4) The trained trackside image intrusion recognition network is used to identify intrusion targets in newly captured trackside monitoring images to obtain the original scale intrusion target recognition results, which include the position, size and category of each intrusion target.

[0037] (1.1.5) Determine whether there is an intrusion target with a size larger than the preset size threshold in the original scale intrusion target identification results. If there is, take the newly captured trackside monitoring image as the trackside monitoring image of the large target and then proceed to step (1.1.6). Otherwise, jump to step (1.1.7).

[0038] (1.1.6) Set the image scaling factor η s Adjust hyperparameters for image resolution; adjust based on image scaling factor. The newly captured trackside monitoring images are scaled down to obtain scaled-down trackside monitoring images. The trained trackside image intrusion recognition network is used to identify intrusion targets in the scaled-down trackside monitoring images to obtain scaled-down intrusion target recognition results. Finally, the scaled-down intrusion target recognition results are joined with the original scaled intrusion target recognition results to obtain the final trackside monitoring foreign object intrusion results.

[0039] (1.1.7) Set the image scaling factor Based on image scaling factor The newly captured trackside monitoring images are reduced in size and enlarged in size to obtain reduced-scale and enlarged-scale trackside monitoring images. The trained trackside image intrusion recognition network is used to identify intrusion targets in the reduced-scale and enlarged-scale trackside monitoring images, respectively, to obtain intrusion target recognition results for both scaled-down and scaled-up images. Finally, the union of the scaled-down intrusion target recognition results, the enlarged-scale intrusion target recognition results, and the original scaled intrusion target recognition results is taken to obtain the final trackside monitoring foreign object intrusion result.

[0040] In this embodiment, when performing target scale-down recognition on trackside monitoring images, the resolution of the scaled-down trackside monitoring image is 1-η times the resolution of the newly captured trackside monitoring image. s When performing target scale magnification recognition on trackside monitoring images, the resolution of the magnified trackside monitoring image is 1 + η times the resolution of the newly captured trackside monitoring image. s By processing foreign object intrusion cases of different scales, we can avoid false detections caused by magnifying larger objects and missed detections of intrusion targets caused by smaller objects.

[0041] (1.2) According to the preset second interval distance, trackside positioning devices and trackside communication devices are installed simultaneously along the subway track line. The trackside positioning devices and trackside calculation devices at the platform, curve and connecting passage are connected to the trackside communication device to transmit the trackside monitoring foreign object intrusion results and line positioning information to the subway line security center.

[0042] In this embodiment, the second interval distance is less than or equal to the maximum communication distance of the trackside communication device. The trackside positioning device is a trackside positioning identification card, and the trackside communication device includes an LET module and a Lora module. The trackside positioning identification card is used to divide the track line sections to accurately provide the positioning information of the railway line sections. Combined with the foreign object intrusion results, the location of the foreign object intrusion can be quickly located, providing a basis for timely clearing of track obstacles and ensuring traffic.

[0043] (1.3) Construct a simulation map of the subway track line based on the subway track line, and map the actual installation locations of the trackside monitoring device, trackside calculation device, trackside positioning device and trackside communication device onto the simulation map of the track line to obtain the subway track line simulation map.

[0044] (2) The image detection device and the radar detection device are set at the front of the subway car to capture real-time visual images of the subway's direction of travel and to detect real-time radar data of the subway's direction of travel.

[0045] In this embodiment, the image detection device uses a visual sensor, and the radar detection device uses millimeter-wave radar; the process of acquiring real-time field-of-view images and real-time radar data is as follows:

[0046] (2.1) The image detection device is set at the center of the front of the subway car, and the image detection device is used to capture real-time visual images of the subway car's direction of travel.

[0047] (2.2) Place the radar detection device above the image detection device so that the radar detection device is located on the central axis of the track line; then construct a rectangular coordinate system with the radar detection device as the origin, the subway travel direction as the longitudinal coordinate axis, and the direction perpendicular to the track line as the transverse coordinate axis.

[0048] (2.3) Based on the difference frequency of the Doppler frequency shift, the radar detection device is used to detect the distance, angle and relative speed of objects in the direction of the subway car's travel until the object is detected.

[0049] In this embodiment, detecting an object means that the radar signal continuously detects an object for more than a preset detection threshold. If the number of times an object is detected is less than the preset detection threshold, the object is considered lost or the detection fails, and the process does not proceed to the next step of processing the detection data. Only when the number of times an object is detected exceeds the preset detection threshold will the next step of processing the detection data begin. In this embodiment, the radar detection device uses a triangular wave modulated signal for foreign object detection.

[0050] (2.4) Number the detected objects and combine the object numbers with the corresponding distance, angle and relative speed to obtain real-time radar data of the subway's direction of travel.

[0051] (3) Use the visual field image foreign object detection neural network in the subway vehicle computing device to identify and detect real-time visual field images, and obtain the visual field image feature detection results;

[0052] (3.1) Obtain several field-view images with or without foreign objects, and perform weighted average grayscale processing and Gaussian filtering processing on the field-view images in sequence to obtain the preprocessed field-view images.

[0053] In this embodiment, the model for weighted average grayscale processing is as follows:

[0054] Gray(a,b)=(ω R R(a,b)+ω G G(a,b)+ω B B(a,b))

[0055] Where Gray(a,b) represents the gray value of the pixel at position (a,b) in the visual field image, ω R ω G ω B R(a,b), G(a,b), and B(a,b) represent the weight coefficients of the red, green, and blue channels, respectively. R(a,b), G(a,b), and B(a,b) represent the red, green, and blue channel pixel values ​​of the pixel at position (a,b) in the visual field image, respectively. a is the horizontal coordinate of the visual field image, and b is the vertical coordinate of the visual field image.

[0056] In this embodiment, the model for Gaussian filtering is as follows:

[0057]

[0058] Where g(a,b) represents the Gaussian filtering result of the pixel at position (a,b) in the grayscale image of the field of view, e represents the exponential basis constant, and σ represents the standard deviation of the Gaussian distribution;

[0059] (3.2) Perform histogram equalization and contrast enhancement on the preprocessed visual field image to obtain the enhanced visual field image;

[0060] (3.3) Use the Sobel operator to perform edge detection on the enhanced field-view image to obtain the edge features of the field-view image;

[0061] In this embodiment, the gradient of the Gaussian-smoothed field of view image in the horizontal direction is calculated using a Sobel operator convolution kernel in the horizontal direction, and the gradient of the Gaussian-smoothed field of view image in the vertical direction is calculated using a Sobel operator convolution kernel in the vertical direction. Then, the edge features of the field of view image are obtained by calculating the magnitude of the gradients in the two directions.

[0062] (3.4) Based on the edge features of the field of view image, Hough transform is used to detect the straight line of the image trajectory and obtain the straight line trajectory features of the field of view image;

[0063] In this embodiment, the Hough transform line detection method is as follows: the field of view image is converted to polar coordinate parameter space and discretized into a grid; based on the edge features of the field of view image, all possible lines corresponding to the pixels with edge features in the field of view image in the parameter space are calculated respectively, and the parameter positions of the lines are accumulated and counted; the pixels whose accumulated count results exceed a set threshold are found, and the line parameters in the field of view image are obtained accordingly, thereby obtaining the orbit features of the field of view image.

[0064] (3.5) Based on the edge features and straight-line trajectory features of the visual field image, the curved trajectory features of the visual field image are obtained through Hough line detection and curve fitting.

[0065] (3.5.1) Based on the edge features of the field of view image, extend along the straight line trajectory features of the field of view image to set a trapezoidal region of interest from the image-enhanced field of view image;

[0066] (3.5.2) Perform perspective transformation on the enhanced field-of-view image and convert it into a bird's-eye view. Based on the straight-line trajectory characteristics of the field-of-view image, use a cubic function to perform curve fitting within the trapezoidal region of interest to obtain the curve trajectory fitting result.

[0067] (3.5.3) If the coefficient of the cubic term in the curve orbit fitting result is greater than... This yields the curve trajectory features of the field-of-view image, where, It is a non-zero constant;

[0068] (3.6) Set the track foreign object intrusion detection limit;

[0069]

[0070] Among them, u ar (v i' ) indicates the right-hand position of the foreign object intrusion detection limit, u ri' U represents a characteristic point on the right side of the subway line. li' U represents a characteristic point on the left side of the subway line. al (v i' ) indicates the left position point of the foreign object detection limit on the track, v i' This represents the horizontal coordinate of the i'th track line;

[0071] (3.7) Replace the standard convolutions in the backbone network and feature fusion network of the traditional yolov11 with depthwise separable convolutions. Then, use a top-down path aggregation mechanism and lateral connections to strengthen the image feature extraction and analysis between layers of the feature fusion network. Use the C2f-Faster-EMA feature fusion strategy to fuse image feature information at different scales to transmit to the neck network, and obtain the improved yolov11 neural network.

[0072] In this embodiment, by replacing the standard convolutions in the backbone and feature fusion networks of YOLOv11 with depthwise separable convolutions, which include both depthwise and pointwise convolutions, feature extraction is performed on each input channel before channel fusion, which can significantly reduce parameters and computational cost without affecting accuracy. In addition, the use of an automatic downward path aggregation mechanism and lateral connections can reduce additional computational overhead. The C2f-Faster-EMA feature fusion strategy can better fuse information at different scales while controlling computational cost.

[0073] (3.8) Based on the foreign object intrusion dataset of the track line, the improved yolov11 neural network is trained for foreign object detection and recognition based on the straight track features of the field of view image, the curved track features of the field of view image, and the foreign object intrusion detection limit of the track, so as to obtain the field of view image foreign object detection neural network.

[0074] In this embodiment, the orbital foreign object intrusion detection boundary can efficiently determine the region of interest during foreign object detection and recognition training for monitoring and recognition training. When training the improved YOLOv11 neural network, half-precision floating-point numbers can be used for gradient updates to reduce memory consumption and improve GPU utilization, thereby accelerating the training speed without affecting convergence.

[0075] (3.9) Use the field of view image foreign object detection neural network to detect real-time field of view images and obtain the field of view image feature detection results.

[0076] (4) Based on the Kalman filter algorithm, the subway onboard computing device is used to identify and monitor the real-time radar data to obtain the current radar detection object position.

[0077] (4.1) Set a continuous monitoring time window for the real-time radar data of the subway's direction of travel, and count the number of times an object is successfully detected within the continuous monitoring time window as the object detection count.

[0078] (4.2) If the number of object detections is less than the first detection threshold, the track line status is determined to be the state of foreign object intrusion disappearance, and return to step (4.1); otherwise, proceed to step (4.3).

[0079] (4.3) If the number of object detections is greater than the second detection threshold, the track line is determined to be in a state of continuous foreign object intrusion and proceeds to step (4.4); otherwise, proceed directly to step (4.1).

[0080] (4.4) Construct a foreign object intrusion state space model based on real-time radar data;

[0081]

[0082] v kτ =v k(τ-1) +a k(τ-1) ×Δt

[0083] Where, p kτ Let v represent the position of the k-th object at time τ. kτ Let a represent the velocity of the k-th object at time τ. k(τ-1) Let Δt represent the acceleration of the k-th object at time τ-1, and let Δt represent the length of the time interval between adjacent objects.

[0084] (4.5) Based on the foreign object intrusion state space model, construct the foreign object intrusion state equation;

[0085]

[0086] Where, ω τ-1 z represents the process noise at time τ-1. τ H represents the position measurement vector at time τ. τ-1 S represents the observation matrix at time τ-1. τ-1 The measurement noise at time τ-1 is represented.

[0087] (4.6) Based on the Kalman filter algorithm, the foreign object intrusion state equation is updated in time and state to obtain the current radar detection object position.

[0088] (5) Based on the data synchronization frequency, the target position is fused by the current radar-detected object position and the field-of-view image feature detection results using the subway vehicle-mounted computing device to obtain the vehicle-mounted intrusion detection results.

[0089] (5.1) The data synchronization frequency is obtained based on the detection frequency of the radar detection device and the detection frequency of the image detection device;

[0090] (5.2) Obtain the transformation model between radar coordinate system and world coordinate system, the transformation model between pixel coordinate system and camera coordinate system, and the transformation model between camera coordinate system and world coordinate system;

[0091] (5.3) Based on the transformation model between the radar coordinate system and the world coordinate system, the current radar detection object position is transformed into the current radar detection position in the world coordinate system;

[0092] (5.4) Based on the transformation model between pixel coordinate system and camera coordinate system and the transformation model between camera coordinate system and world coordinate system, the field image feature detection results are converted into image detection results in world coordinate system;

[0093] (5.5) Based on the data synchronization frequency, the target position is fused with the current radar detection position and image detection result at the same time according to the following target association model to obtain the vehicle intrusion detection result;

[0094] The target association model is as follows:

[0095]

[0096] Where Δd represents the distance between the current radar detection position and the center point of the object in the image detection result, O R Indicates the current radar detection location, O C The image detection result indicates the location of the object's center point, l1 represents the first correlation threshold, IOU represents the intersection-union ratio of the radar and image-detected object regions, IOU represents the contour region of the object detected by the radar, and S represents the object's contour region detected by the radar. C l1 represents the contour region of the object being detected in the image, and l2 represents the second correlation threshold.

[0097] In this embodiment, if the distance between the current radar detection position and the center point of the object in the image detection result is less than the first correlation threshold, and the cross-union ratio of the radar and the object detection area in the image is greater than the second correlation threshold, the object fused at the target position can be effectively used as the vehicle intrusion detection result.

[0098] (6) The subway onboard computing device combines the onboard intrusion detection results and the trackside monitoring foreign object intrusion results to map and alarm on the subway track line simulation map.

[0099] (6.1) Upload the intrusion detection results to the subway line security center through the subway vehicle communication device;

[0100] (6.2) Map the results of foreign object intrusion from trackside monitoring and vehicle-mounted intrusion detection onto the subway track line simulation map to generate a track foreign object intrusion warning map.

[0101] (6.3) Send a warning map of foreign object intrusion warning to each subway vehicle communication device through the subway line security center.

[0102] As shown in Figure 2, this invention provides a system for identifying foreign object intrusions in subway lines based on the aforementioned method, namely, a subway line foreign object intrusion monitoring system. The subway line foreign object intrusion detection system includes several trackside foreign object intrusion detection subsystems, a subway line security center, and onboard foreign object intrusion monitoring subsystems sequentially installed along subway tracks at platforms, curves, and connecting passages. The trackside foreign object intrusion detection subsystem can continuously monitor and identify areas prone to foreign object intrusions without blind spots, and locate and upload the trackside monitoring results to the subway line security center. The onboard foreign object intrusion monitoring subsystem can detect and identify foreign object intrusions by combining images and radar, and upload the onboard intrusion detection results in the direction of subway travel to the subway line security center. The subway line security center maps the trackside monitoring results and the onboard intrusion detection results onto a simulated subway track map, generating a track foreign object intrusion warning map, and sends the track foreign object intrusion warning map to each subway onboard communication device for alarm purposes.

[0103] This invention comprehensively utilizes technologies such as trackside monitoring, on-board monitoring, neural network recognition, Kalman filtering algorithm, and visual management to fully enhance the detection capability, real-time performance and accuracy, intelligent fusion and target positioning capability, as well as visual management and rapid response capability of foreign object intrusion in subways, providing strong protection for subway operation safety.

[0104] Although the illustrative specific embodiments of the present invention have been described above to enable those skilled in the art to understand the invention, it should be understood that the invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the invention as defined and determined by the appended claims, and all inventions utilizing the concept of the present invention are protected.

Claims

1. A method for identifying foreign object intrusion in subway lines, characterized in that, The steps include: (1) setting up trackside monitoring, positioning, calculation and communication devices along the subway track line to identify and upload the results of foreign object intrusion monitoring along the trackside, and to construct a simulated map of the subway track line; (2) The image detection device and the radar detection device are installed at the front of the subway car to capture real-time visual images of the subway's direction of travel and to detect real-time radar data of the subway's direction of travel. (3) Use the foreign object detection neural network in the field of view image of the subway vehicle computing device to identify and detect the real-time field of view image and obtain the field of view image feature detection results; (3.1) Obtain several field-view images with or without foreign objects, and perform weighted average grayscale processing and Gaussian filtering processing on the field-view images in sequence to obtain preprocessed field-view images; (3.2) Perform histogram equalization processing and contrast enhancement processing on the preprocessed field-view images to obtain enhanced field-view images; (3.3) Use the Sobel operator to perform edge detection on the enhanced field-view images to obtain the edge features of the field-view images; (3.4) Based on the edge features of the field of view image, Hough transform is used to detect the straight line of the image trajectory to obtain the straight line trajectory features of the field of view image; (3.5) Based on the edge features and straight-line trajectory features of the visual field image, the curved trajectory features of the visual field image are obtained through Hough line detection and curve fitting. (3.6) Set the track foreign object intrusion detection limit; ;in, This indicates the right-hand position of the foreign object intrusion detection clearance. This indicates the characteristic points of the right track of the subway line. This indicates the characteristic points of the left track of the subway line. This indicates the left-hand position of the foreign object detection clearance on the track. Indicates the first (3.7) Replace the standard convolutions in the backbone network and feature fusion network of the traditional YOLOv11 with depthwise separable convolutions. Then, use a top-down path aggregation mechanism and lateral connections to strengthen the image feature extraction and analysis between layers of the feature fusion network. Use the C2f-Faster-EMA feature fusion strategy to fuse image feature information at different scales to transmit to the neck network to obtain the improved YOLOv11 neural network. (3.8) Based on the foreign object intrusion dataset of the track line, perform foreign object detection on the improved YOLOv11 neural network based on the straight track features of the field of view image, the curved track features of the field of view image, and the foreign object intrusion detection limit of the track. (3.9) The visual field image foreign object detection neural network is obtained by training the visual field image foreign object detection neural network; (4) The visual field image feature detection result is obtained by using the visual field image foreign object detection neural network to detect the real-time visual field image; (5) According to the Kalman filter algorithm, the subway vehicle-mounted computing device is used to identify and monitor the real-time radar data to obtain the current radar detection object position; (6) According to the data synchronization frequency, the subway vehicle-mounted computing device is used to fuse the current radar detection object position and the visual field image feature detection result to obtain the vehicle-mounted intrusion detection result; (7) The subway vehicle-mounted computing device combines the vehicle-mounted intrusion detection result and the trackside monitoring foreign object intrusion result to perform mapping alarm on the subway track line simulation map.

2. The method for identifying foreign object intrusion in subway lines according to claim 1, characterized in that, Step (1) includes the following steps: (1.1) According to the preset first interval distance, trackside monitoring devices and trackside computing devices connected to the trackside monitoring devices are set up sequentially along the subway track line at the platform, curve and connecting passage. The trackside monitoring devices are used to capture trackside monitoring images; the trackside computing devices are used to identify the foreign object intrusion results in the trackside monitoring images; (1.2) According to the preset second interval distance, trackside positioning devices and trackside communication devices are set up sequentially along the subway track line. The trackside positioning devices and trackside computing devices at the platform, curve and connecting passage are all connected to the trackside communication devices to transmit the trackside monitoring foreign object intrusion results combined with the line positioning information to the subway line security center; (1.3) A line simulation map is constructed according to the subway track line, and the actual setting positions of the trackside monitoring devices, trackside computing devices, trackside positioning devices and trackside communication devices are mapped to the line simulation map to obtain the subway track line simulation map.

3. The method for identifying foreign object intrusion in subway lines according to claim 2, characterized in that, The method for identifying the foreign object intrusion results in the trackside monitoring images in step (1.1) is as follows: (1.1.1) Acquire several trackside monitoring images to form a trackside monitoring image dataset; (1.1.2) Keep the backbone and neck network of the YOLOv5 network unchanged, replace the three prediction head sub-networks in the prediction head network with a classification prediction sub-network, a size prediction sub-network, and an offset prediction sub-network to obtain a trackside image intrusion identification network, and then set the trackside image intrusion identification network in the trackside computing device; (1.1.3) According to the multi-scale target loss function, input the trackside monitoring images into the trackside image intrusion identification network for intrusion target identification prediction training to obtain the trained trackside image intrusion identification network; (1.1.4) Using the trained trackside image intrusion recognition network, intrusion targets are identified in the newly captured trackside monitoring images to obtain the original scale intrusion target recognition results, which include the position, size, and category of each intrusion target; (1.1.5) It is determined whether there are intrusion targets with a size larger than a preset size threshold in the original scale intrusion target recognition results. If so, the newly captured trackside monitoring image is used as a large target trackside monitoring image, and then proceed to step (1.1.6). Otherwise, proceed to step (1.1.7); (1.1.6) Set the image scaling factor. ， Adjust hyperparameters for image resolution; adjust based on image scaling factor. The newly captured trackside monitoring images are scaled down to obtain scaled-down trackside monitoring images; the trained trackside image intrusion recognition network is used to identify intrusion targets in the scaled-down trackside monitoring images to obtain scaled-down intrusion target recognition results; finally, the scaled-down intrusion target recognition results are joined with the original scaled intrusion target recognition results to obtain the final trackside monitoring foreign object intrusion results; (1.1.7) Set the image scaling factor 、 According to the image scaling factor 、 The newly captured trackside monitoring images are reduced in size and enlarged in size to obtain reduced-scale and enlarged-scale trackside monitoring images. The trained trackside image intrusion recognition network is used to identify intrusion targets in the reduced-scale and enlarged-scale trackside monitoring images, respectively, to obtain intrusion target recognition results for both reduced-scale and enlarged-scale images. Finally, the union of the reduced-scale, enlarged-scale, and original-scale intrusion target recognition results is taken to obtain the final trackside monitoring foreign object intrusion result.

4. The method for identifying foreign object intrusion in subway lines according to claim 3, characterized in that, The multi-scale objective loss function is: ； ； ； ;in, Represents a multi-scale objective loss function. Represents the classification loss function. This represents the hyperparameters of the first loss function. This represents the offset loss function. This represents the hyperparameters of the second loss function. Let represent the size loss function, N represent the total number of intruding targets, C represent the number of channels, H represent the height of the trackside monitoring image, W represent the width of the trackside monitoring image, c represent the category of pixels in the image, y represent the ordinate of a pixel in the image, and x represent the abscissa of a pixel in the image. This represents the probability that a pixel belongs to the center point of a certain category. This represents the true probability that a pixel belongs to the center point of a certain category. This represents the first category index factor. This represents the second category index factor. This represents the offset between the center point of the i-th intruding target in the trackside monitoring image and the center point of the actual target bounding box in the trackside monitoring image. This represents the true value of the offset between the center point of the i-th intruding target in the trackside monitoring image and the center point of the actual target bounding box in the trackside monitoring image. This indicates the search for the first norm. This indicates the size of the i-th intrusion target corresponding to the predicted center point. This represents the actual size of the i-th intrusion target.

5. The method for identifying foreign object intrusion in subway lines according to claim 1, characterized in that, The step (2) includes the following steps: (2.1) The image detection device is set at the center of the front of the subway car, and the image detection device is used to capture a real-time field-of-view image of the subway car's direction of travel; (2.2) Place the radar detection device above the image detection device so that the radar detection device is located on the central axis of the track line; then construct a rectangular coordinate system with the radar detection device as the origin, the subway travel direction as the longitudinal coordinate axis, and the direction perpendicular to the track line as the transverse coordinate axis. (2.3) Based on the difference frequency of the Doppler frequency shift, the radar detection device is used to detect the distance, angle and relative speed of objects in the direction of the subway car's travel until the object is detected; (2.4) Number the detected objects and combine the object number with the corresponding distance, angle and relative speed to obtain real-time radar data of the subway's direction of travel.

6. The method for identifying foreign object intrusion in subway lines according to claim 1, characterized in that, The step (4) includes the following steps: (4.1) Set a continuous monitoring time window for the real-time radar data of the subway travel direction, and count the number of times the object is successfully detected within the continuous monitoring time window as the object detection count; (4.2) If the number of object detections is less than the first detection threshold, the track line status is determined to be the foreign object intrusion disappearance state, and return to step (4.1); otherwise, proceed to step (4.3); (4.3) If the number of object detections is greater than the second detection threshold, the track line status is determined to be the foreign object intrusion continuous state, and proceed to step (4.4); otherwise, proceed directly to step (4.1); (4.4) Construct a foreign object intrusion state space model based on real-time radar data; ； ;in, Indicates the k-th object in Location at any given moment Indicates the k-th object in The speed of time, Indicates the k-th object in acceleration at any moment Represents the length of adjacent time intervals; (4.5) Based on the foreign object intrusion state space model, construct the foreign object intrusion state equation; ； ;in, express Time-based process noise, express Position measurement vector at time, express The observation matrix at time, express Measurement noise at any time; (4.6) According to the Kalman filter algorithm, the foreign object intrusion state equation is updated in time and state to obtain the current radar detection object position.

7. The method for identifying foreign object intrusion in subway lines according to claim 1, characterized in that, Step (5) includes the following steps: (5.1) Obtain the data synchronization frequency based on the detection frequency of the radar detection device and the detection frequency of the image detection device; (5.2) Obtain the transformation model between the radar coordinate system and the world coordinate system, the transformation model between the pixel coordinate system and the camera coordinate system, and the transformation model between the camera coordinate system and the world coordinate system; (5.3) Based on the transformation model between the radar coordinate system and the world coordinate system, convert the current radar detection object position into the current radar detection position in the world coordinate system. (5.4) Based on the transformation model between the pixel coordinate system and the camera coordinate system and the transformation model between the camera coordinate system and the world coordinate system, the field-of-view image feature detection results are converted into image detection results in the world coordinate system; (5.5) According to the data synchronization frequency, the target position is fused with the current radar detection position and image detection results at the same time based on the following target association model to obtain the vehicle intrusion detection results; wherein, the target association model is: ;in, This indicates the distance between the current radar detection location and the center point of the object in the image detection result. Indicates the current radar detection location. This indicates the location of the center point of the object in the image detection results. This represents the first correlation threshold. This represents the cross-over ratio (COP) between the radar and the image-detected object regions. This indicates the outline region of an object detected by radar. This represents the outline region of the object detected in the image. This represents the second correlation threshold.

8. The method for identifying foreign object intrusion in subway lines according to claim 1, characterized in that, The step (6) includes the following steps: (6.1) Uploading the vehicle intrusion detection results to the subway line security center through the subway vehicle communication device; (6.2) Map the results of foreign object intrusion monitoring by the trackside and the results of vehicle-mounted intrusion detection onto the subway track line simulation map to generate a track foreign object intrusion warning map; (6.3) The subway security center sends a warning map of foreign object intrusion into the track to each subway vehicle communication device to issue an alarm.

Images