Track fastener installation state detection method and system

Abstract

The invention discloses a track fastener installation state detection method and system. The method comprises the steps that fastener images are dynamically collected along a track, the collected fastener images are synchronized with time or space, and continuous fastener images are obtained; using a pre-trained YOLO-OBB detection model to identify the position of an elastic strip in each fastener image, and marking an orientation angle based on an elastic strip OBB frame to obtain an angle sequence of continuous fasteners; and detecting sudden change of the angle of the adjacent fasteners, and when the angle exceeds a preset threshold value, starting local area analysis to confirm the reverse installation state. According to the method, the reversely-installed fastener can be accurately identified, and the precision ratio and recall ratio of detection are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of rail transit technology, and more specifically to a method and system for detecting the installation status of rail fasteners. Background Technology

[0002] Track fasteners are key components in the track structure, used to ensure a secure connection between the rails and sleepers, prevent longitudinal and lateral movement of the rails, reduce track vibration, and adjust the track gauge.

[0003] During track laying and maintenance, the problem of fasteners being installed backwards may occur. Reverse installation of fasteners can reduce track structural stability, increase train operation risks, and may even lead to derailment accidents.

[0004] Currently, the inspection of fastener installations mostly relies on manual inspections. This method is not only inefficient but also easily affected by subjective factors, making it difficult to guarantee the accuracy and consistency of the inspections. Furthermore, the effectiveness of manual inspections is limited under certain environmental conditions (such as at night or inside tunnels). With the rapid development of rail transit networks and the continuous increase in operating mileage, relying solely on manual inspections is insufficient to achieve timely and comprehensive coverage of the entire line, and it is also difficult to guarantee the accuracy of the inspections.

[0005] In contrast, automated inspection not only improves inspection efficiency but also ensures the objectivity and reliability of inspection results. However, how to apply it to the identification of reverse fasteners is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] In view of this, in order to at least partially solve the above-mentioned technical problems, the present invention provides a method and system for detecting the installation status of track fasteners based on computer vision, aiming to improve the detection efficiency and accuracy of track fastener installation status.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] Firstly, this application discloses a method for detecting the installation status of track fasteners, including,

[0009] Dynamically acquire images of fasteners along the track;

[0010] Synchronize the acquired fastener images with time or space to obtain continuous fastener images;

[0011] The pre-trained YOLO-OBB detection model is used to identify the position of the spring bar in each fastener image, and the orientation angle is marked based on the OBB box of the spring bar to obtain the angle sequence of continuous fasteners;

[0012] Detect sudden changes in the angle of adjacent fasteners. When the angle exceeds a preset threshold, initiate local area analysis to confirm the reverse installation status.

[0013] In one optional embodiment, the acquired fastener image is synchronized with the vehicle odometer in time and / or with GPS location information in space.

[0014] In one optional embodiment, the training data of the YOLO-OBB detection model includes OBB labeled samples of fasteners installed normally and in reverse in straight / curved rail scenarios.

[0015] In one optional embodiment, detecting abrupt changes in the angle of adjacent fasteners includes the following steps:

[0016] If the angle change exceeds the preset threshold, the reverse installation status is confirmed by local area analysis. If it is, it is marked as reverse installation; otherwise, it is not marked or is marked as normal installation.

[0017] If the angle change does not exceed the preset threshold, it will not be marked or will be marked as normal installation.

[0018] In one optional embodiment, local region analysis refers to verifying the angular consistency of each of the m (m≥3) fasteners before and after the point of change.

[0019] In one optional embodiment, the mutation detection results are visualized and output, including fastener position, angle, and reverse mounting mark.

[0020] Secondly, this application discloses a track fastener installation status detection system, comprising:

[0021] The image acquisition unit includes at least five anti-vibration high-definition track cameras that are evenly distributed laterally on the bottom of the inspection vehicle. These cameras are used to dynamically acquire images of the fasteners along the track according to the vehicle speed. The acquired fastener images are synchronized with time or space to obtain continuous fastener images.

[0022] The image detection unit is used to identify the position of the spring bar in each fastener image using a pre-trained YOLO-OBB detection model, and to obtain the angle sequence of continuous fasteners based on the orientation angle marked by the OBB box of the spring bar.

[0023] The reverse installation judgment unit is used to detect sudden changes in the angle of adjacent fasteners. When the angle exceeds a preset threshold, local area analysis is initiated to confirm the reverse installation status.

[0024] In one optional embodiment, local region analysis refers to verifying the angular consistency of m fasteners before and after the point of change.

[0025] In one optional embodiment, a result output unit is further included for displaying the fastener position, angle, and reverse mounting mark.

[0026] Compared with existing technologies, the track fastener installation status detection method and system of the present invention can accurately identify reverse-installed fasteners by analyzing the orientation angle of the fastener spring strip, significantly improving the accuracy and recall rate of detection, thereby providing a strong guarantee for the safe operation of rail transit. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0028] Figure 1 Flowchart of the track fastener installation status detection method provided by the present invention;

[0029] Figure 2 This is a structural example diagram of the track fastener installation status detection system provided by the present invention. Detailed Implementation

[0030] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0032] This invention discloses a method and system for detecting the installation status of track fasteners based on computer vision. Visual data can intuitively reflect the actual installation status of the fasteners and is easier to judge and analyze than data from other sensors.

[0033] The following is a description through specific embodiments.

[0034] Example 1:

[0035] The method for detecting the installation status of track fasteners in this embodiment includes the following steps:

[0036] Dynamically acquire images of fasteners along the track;

[0037] Synchronize the acquired fastener images with time or space to obtain continuous fastener images;

[0038] The pre-trained YOLO-OBB detection model is used to identify the position of the spring bar in each fastener image, and the orientation angle is marked based on the OBB box of the spring bar to obtain the angle sequence of continuous fasteners;

[0039] Detect sudden changes in the angle of adjacent fasteners. When the angle exceeds a preset threshold, initiate local area analysis to confirm the reverse installation status.

[0040] In one embodiment, cameras are deployed at the bottom of the inspection vehicle to dynamically acquire images of fasteners as the vehicle moves along. Preferably, multiple cameras are deployed at the bottom of the vehicle in a direction perpendicular to the track to achieve synchronous inspection of the left and right tracks.

[0041] In this embodiment, the collected fastener images are synchronized with the vehicle odometer in time and / or with GPS location information in space to obtain continuous fastener images, which facilitates subsequent reverse installation detection.

[0042] The YOLO-OBB detection model's network architecture combines a YOLO backbone network with an OBB detection head, fully leveraging the efficiency of the YOLO series models. Simultaneously, the OBB detection head enables accurate estimation of the fastener's orientation. The Oriented Bounding Box (OBB) format is used to annotate the elastic bar's position and orientation, accurately capturing its spatial orientation and effectively handling fastener attitude changes in complex scenarios such as curves.

[0043] During model training, OBB-annotated samples containing both normal and reverse fastener installations under straight / curved rail scenarios were selected to increase the diversity of the dataset and improve the model's adaptability to different rail environments. Furthermore, the image data was augmented using arbitrary 360-degree rotations, enabling the model to learn the feature representations of the fasteners at any angle.

[0044] In one embodiment, the detection model is trained using the deep learning framework PyTorch. The training process employs a cosine annealing learning rate strategy, lasting for 200 epochs, and the model performance is periodically evaluated on a validation set. After training, the model is comprehensively evaluated using a test set to ensure its effectiveness in practical applications.

[0045] Furthermore, this application innovatively proposes a reverse-installation determination method based on abrupt changes in orientation angle. By checking the orientation angle of the fastener spring along the track, reverse-installation fasteners can be effectively identified. The specific process is as follows: Figure 1 ,include:

[0046] Fastener area inspection, fastener spring strip positioning;

[0047] Calculate the orientation angle of each detected fastener spring bar, ranging from 0° to 360°;

[0048] The orientation angles of the continuous fastener spring strips are sorted along the track direction;

[0049] Calculate the difference in the orientation angle of adjacent fastener spring strips to detect abrupt changes;

[0050] When the detected orientation angle difference is close to the preset threshold (considering a certain tolerance range, such as 175° to 185°), it is determined that there may be a reverse fastener; otherwise, it is marked as normal installation.

[0051] To improve reliability, this application considers the orientation angle distribution of multiple fasteners within a local area and identifies anomalies through statistical analysis, thereby reducing the probability of misjudgment caused by detection errors of individual fasteners. When an anomaly is identified, it is marked as reverse installation. Otherwise, it is not marked or marked as normal installation.

[0052] The fastener installation status detection method of this application makes full use of the spatial continuity characteristics of fastener installation, which significantly improves the accuracy of detecting reverse-installed fasteners.

[0053] Example 2:

[0054] This application presents a track fastener installation status detection system that automates the entire process from data acquisition to result output. The system can be installed on existing track inspection vehicles and works in conjunction with other inspection systems to provide comprehensive track condition monitoring data.

[0055] In this embodiment, the track fastener installation status detection system includes:

[0056] The image acquisition module includes at least five anti-vibration high-definition track cameras that are evenly distributed laterally on the bottom of the inspection vehicle. These cameras are used to acquire images of the track fasteners in real time. The acquisition frequency is automatically adjusted according to the vehicle speed to ensure the overlap between images. The acquired fastener images are synchronized with time (from the vehicle odometer) or space (GPS system) to obtain continuous fastener images.

[0057] The arrangement of five cameras achieves full coverage of the track section, including the rails on both sides and the central area, effectively avoiding missed and false detections. Furthermore, the high-resolution cameras can capture subtle features of the fasteners, providing high-quality raw data for subsequent image analysis and processing.

[0058] In one embodiment, the camera employs an advanced optical system and image sensor to ensure image clarity and stability during high-speed movement; it also features dustproof and shockproof design, as well as automatic exposure and white balance adjustment functions to ensure all-weather, all-terrain data acquisition capabilities.

[0059] As a preferred method, all cameras are triggered and data is acquired synchronously to simultaneously detect and display all fasteners on the track section.

[0060] Compared to manual data collection, this method can collect data continuously at normal train speeds, significantly improving data acquisition efficiency.

[0061] After acquiring the image, this embodiment uses a high-speed data transmission and storage system, combined with an advanced data compression algorithm, to transmit the data to the image processing unit in real time.

[0062] The image detection unit is used to identify the position of the spring bar in each fastener image using a pre-trained YOLO-OBB detection model, and to obtain the angle sequence of continuous fasteners based on the orientation angle marked by the OBB box of the spring bar.

[0063] The reverse installation judgment unit is used to detect sudden changes in the angle of adjacent fasteners. When the angle exceeds a preset threshold, local area analysis is initiated to confirm the reverse installation status.

[0064] In one optional embodiment, local region analysis refers to verifying the angular consistency of m fasteners before and after the point of change.

[0065] In one optional embodiment, a result output unit is further included for displaying the fastener position, angle, and reverse mounting mark.

[0066] In one exemplary application, the track fastener installation status detection system refers to... Figure 2 It mainly consists of three subsystems: a data acquisition subsystem, an image processing subsystem, and a decision-making and output subsystem; among which,

[0067] The data acquisition subsystem includes a high-definition camera array and a position synchronization module. The position synchronization module is connected to the vehicle odometer and / or GPS system and is used to perform spatiotemporal alignment on the fastener images acquired by the high-definition cameras. After alignment, the images are sent to the image acquisition module.

[0068] The image processing subsystem includes, in sequence, an image preprocessing module (which performs noise reduction, contrast enhancement, and other processing on the original image to improve image quality), a fastener detection module, and a direction estimation module. The execution principle of each module is as described in Example 1, and will not be repeated here.

[0069] The decision-making and output subsystem includes a reverse-installation judgment module for identifying reverse-installed fasteners. To further optimize the solution, a result visualization module, an alarm module, and a data storage module are added, all of which are connected to a communication interface. The result visualization module includes an on-board display and a wireless communication module, installed in the driver's cab, facilitating real-time monitoring of the system's operating status and detection results by the operator. The displayed content includes track images, the location of detected fasteners, the orientation angle of each fastener, and markings of fasteners determined to be reverse-installed. This information intuitively demonstrates the detection process and results, enabling operators to quickly identify problem areas.

[0070] Through this system, the present invention realizes automated, efficient, and high-precision detection of track fastener reverse installation, providing a strong guarantee for the safe operation of rail transit.

[0071] In another embodiment, this embodiment uses a low-power edge computing device for algorithm deployment and real-time processing, thereby achieving a miniaturized design. Specifically, it includes a dedicated cabinet installed inside the carriage, which is connected to the image acquisition unit via high-speed Ethernet to achieve real-time data transmission.

[0072] Furthermore, by optimizing hardware design and heat dissipation solutions, the device height is less than 2U, facilitating installation and maintenance; a fanless cooling design ensures stable operation over a wide temperature range; and special vibration damping design and solid-state storage effectively address vibrations and impacts encountered during vehicle operation. In addition, a series of optimization strategies were employed, including model quantization, distillation, memory optimization, parallel computing, and pipelined processing, to achieve efficient fastener reverse assembly detection with limited computing resources.

[0073] Through the above optimizations, this application achieves real-time processing of high-definition images on edge computing devices, reaching a processing speed of 60 frames per second, while maintaining a high detection accuracy.

[0074] The entire system in this application is powered by the vehicle's auxiliary power system, with voltage regulators and uninterruptible power supplies ensuring the stability and reliability of the power supply. All signal and power lines between components are designed to resist interference and are routed through dedicated cable trays, minimizing the risk of electromagnetic interference and mechanical damage, thus providing crucial protection for rail transit safety.

[0075] Compared with the prior art, this application has the following advantages:

[0076] 1) High-precision recognition and strong adaptability. This invention uses a deep learning model combined with OBB detection technology to accurately capture the spatial orientation of the fastener spring. Through an innovative orientation angle abrupt change detection algorithm, it can effectively identify reverse-installed fasteners and adapt to different scenarios such as straight and curved rails. The application of data augmentation strategies further improves the model's ability to recognize fasteners at various angles.

[0077] 2) Real-time, efficient, and easy to integrate. This invention employs an optimized lightweight deep learning model and efficient edge computing devices to achieve real-time processing of high-resolution images. The modular system design makes this device easy to integrate with existing track inspection vehicles and can be deployed without large-scale modifications.

[0078] 3) High degree of automation. This application features comprehensive automation design, which significantly reduces manual intervention, improves detection efficiency and consistency, and enables large-scale, long-term track monitoring.

[0079] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0080] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for detecting the installation status of track fasteners, characterized in that, include: Images of fasteners are dynamically acquired along the track and synchronized with time or space to obtain continuous images of fasteners. The pre-trained YOLO-OBB detection model is used to identify the position of the spring bar in each fastener image, and the orientation angle is marked based on the OBB box of the spring bar to obtain the angle sequence of continuous fasteners; Detect sudden changes in the angle of adjacent fasteners. When the angle exceeds a preset threshold, initiate local area analysis to confirm the reverse installation status.

2. The method for detecting the installation status of track fasteners according to claim 1, characterized in that, The collected fastener images are synchronized with the vehicle odometer in time and / or with GPS location information in space.

3. The method for detecting the installation status of track fasteners according to claim 1, characterized in that, The training data for the YOLO-OBB detection model includes OBB labeled samples of fasteners installed normally and in reverse in straight / curved rail scenarios.

4. The method for detecting the installation status of track fasteners according to claim 1, characterized in that, The steps for detecting abrupt changes in the angle of adjacent fasteners include: If the angle change exceeds the preset threshold, the reverse installation status is confirmed by local area analysis. If it is, it is marked as reverse installation; otherwise, it is not marked or is marked as normal installation. If the angle change does not exceed the preset threshold, it will not be marked or will be marked as normal installation.

5. The method for detecting the installation status of track fasteners according to claim 1 or 4, characterized in that, Local area analysis refers to verifying the angular consistency of m fasteners before and after the point of change.

6. The method for detecting the installation status of track fasteners according to claim 1, characterized in that, The mutation detection results are visualized, including the fastener position, angle, and reverse mounting mark.

7. A track fastener installation status detection system, characterized in that, include: The image acquisition unit includes at least five anti-vibration high-definition track cameras that are evenly distributed laterally on the bottom of the inspection vehicle. These cameras are used to dynamically acquire images of the fasteners along the track according to the vehicle speed. The acquired fastener images are synchronized with time or space to obtain continuous fastener images. The image detection unit is used to identify the position of the spring bar in each fastener image using a pre-trained YOLO-OBB detection model, and to obtain the angle sequence of continuous fasteners based on the orientation angle marked by the OBB box of the spring bar. The reverse installation judgment unit is used to detect sudden changes in the angle of adjacent fasteners. When the angle exceeds a preset threshold, local area analysis is initiated to confirm the reverse installation status.

8. The track fastener installation status detection system according to claim 7, characterized in that, Local area analysis refers to verifying the angular consistency of m fasteners before and after the point of change.

9. The track fastener installation status detection system according to claim 7, characterized in that, It also includes a result output unit for displaying the fastener position, angle, and reverse mounting mark.

Images