Two-stage tunnel apparent communication signal cable shedding detection method and system

Abstract

The invention relates to a two-stage tunnel appearance communication signal cable falling detection method and system. The detection method comprises the following steps: S1, dynamically acquiring an original image of tunnel appearance ancillary facility equipment by using an image acquisition module; s2, preprocessing the original image acquired by the image acquisition module; s3, classifying tunnel appearance ancillary facilities and equipment based on the preprocessed image data; s4, the slope of the communication signal cable is detected, and falling-off detection of the communication signal cable is achieved. The detection system comprises an image acquisition module which is used for dynamically acquiring an original image of tunnel appearance ancillary facility equipment; and the processing module is used for pre-processing the original image acquired by the image acquisition module, classifying tunnel appearance ancillary facilities and equipment based on the pre-processed image data, and detecting the slope of the communication signal cable so as to realize fall-off detection of the communication signal cable.

Description

Technical Field

[0001] The present invention relates to the field of urban rail transit communication technology, and in particular to a method and system for detecting the detachment of a second-stage tunnel apparent communication signal cable. Background Art

[0002] As the "central nervous system" of urban rail transit, the communication signal system plays an irreplaceable role in ensuring driving safety, improving transportation efficiency and optimizing driving organization. With the continuous improvement of the operating speed of urban rail transit (urban rail) and the increase in the service life of infrastructure equipment, the problems of urban rail equipment failure and facility aging have gradually emerged. The frequent operational failures and rising maintenance costs caused by this have brought severe challenges to the safety and smooth operation of urban rail transit. Especially for those communication signal cables that have been exposed to complex environments for a long time, their health status is directly related to the accurate transmission of train control information. Any damage or shedding may cause serious safety hazards, and even lead to limit violations, threatening the safety of passengers' lives and property. Therefore, regular inspection, monitoring and maintenance of these key facilities have become an important task to ensure the stable operation of the urban rail system.

[0003] CN104410820A discloses a vehicle-mounted trackside equipment box and cable appearance detection system. It includes an image acquisition device, which is installed at a vehicle body position corresponding to the upper and oblique outer side of the track, and is used to collect image data of the trackside equipment box and cable in real time according to a control signal of a controller; a controller, which is used to send control signals to a plurality of image acquisition devices according to image acquisition parameters input by a storage display server; and a storage display server, which is used to obtain environmental information from an environmental information acquisition device, and obtain image acquisition parameters matching the environmental information according to the environmental information.

[0004] CN112200483A discloses an automatic inspection system and method for subway trackside equipment. An image acquisition module, a positioning module, and a fault detection module are arranged on the inspection vehicle. The image acquisition module acquires image data of the subway trackside equipment, the positioning module acquires the real-time position information of the inspection vehicle, and the fault detection module identifies and analyzes the subway trackside equipment based on the image data and the corresponding position information when the image is taken, acquires the position of the faulty equipment and specifies the fault type.

[0005] CN111208146A discloses a tunnel cable detection system and detection method, the system includes: an image acquisition device arranged outside the inspection vehicle compartment, a speed sensor arranged on the inspection vehicle wheel, and an image processing device connected to the image acquisition device; the image acquisition device is used to acquire image information of the cable to be detected, and send the image information to the image processing device; the image processing device includes an image data acquisition control module and an image data processing module, the image data acquisition control module includes a synchronous acquisition control submodule, which is used to trigger the image acquisition device to acquire image information according to the speed information of the inspection vehicle wheel acquired by the speed sensor; the image data processing module is used to process the image information. The method is a method for tunnel cable detection using the above system.

[0006] The existing methods for detecting damage and detachment of communication signal cables mainly rely on manual inspection or the detection system equipped with a low-speed electric drive integrated detection platform. The former has obvious limitations because it requires staff to personally enter the track area for visual inspection: on the one hand, the efficiency of manual inspection is extremely low, especially in long-distance lines, and it is time-consuming and laborious to complete a comprehensive inspection; on the other hand, this method is subject to the technical level and personal experience of the inspectors, and the detection results between different personnel may vary greatly, and potential hidden dangers are easily missed. In addition, manual inspection is also accompanied by higher safety risks, especially at night or in bad weather conditions, the inspectors face greater physical and psychological pressure. Although the latter uses mechanized means, it still has many shortcomings. For example, the speed of detection using an electric drive integrated detection platform is usually limited to about 15km / h, which is much lower than the actual train running speed, so that such detection can only be arranged during non-operating hours, which greatly limits the detection frequency and timeliness. At the same time, the detection accuracy and reliability of such systems are also subject to certain limitations, especially in the face of complex and changeable tunnel environments, the accuracy of the final results may be affected by factors such as lighting conditions and background interference. In summary, both traditional and currently more advanced detection technologies have shown varying degrees of lag and limitations in responding to the rapidly developing needs of urban rail transit, and there is an urgent need to explore a new generation of detection solutions that are more efficient, accurate, and safe.

[0007] In addition, on the one hand, there are differences in understanding among those skilled in the art; on the other hand, the applicant studied a large number of documents and patents when making the present invention, but due to space limitations, not all details and contents are listed in detail. However, this does not mean that the present invention does not have the characteristics of these prior arts. On the contrary, the present invention already has all the characteristics of the prior art, and the applicant reserves the right to add relevant prior art to the background technology. Summary of the invention

[0008] In view of the deficiencies in the prior art, the present invention provides a method and system for detecting the detachment of a two-stage tunnel surface communication signal cable to solve at least some of the above-mentioned technical problems.

[0009] In the present invention, "two stages" refers to the two main steps included in the detection method. First, the image is classified and identified, and then the slope of the communication signal cable is specifically detected to determine whether it has fallen off. Furthermore, "two stages" specifically refers to a systematic, step-by-step detection process. The first stage involves preprocessing and classifying the collected images of equipment inside the tunnel to ensure that various types of ancillary facilities and equipment can be accurately identified; the second stage focuses on the status analysis of the communication signal cables, especially by calculating the slope of the cables to assess whether there is a risk of falling off. These two stages together constitute a complete automated detection solution aimed at improving the safety and reliability of urban rail transit communication systems.

[0010] In the present invention, "tunnel appearance" refers to the visible surface inside the tunnel and the various equipment and structures installed on it. Furthermore, "tunnel appearance" specifically refers to the part of the tunnel that can be directly observed, including but not limited to the tunnel walls, ceiling and all ancillary facilities and equipment fixed thereon, such as junction boxes, broadcast speakers, communication signal cables, cable brackets, etc. These elements constitute part of the tunnel's visual environment, and their status is crucial to ensuring the safety of train operation. Therefore, when performing communication signal cable detachment detection, it is necessary to fully consider the entire tunnel appearance to obtain the most accurate results.

[0011] In the present invention, "communication signal cable" is a key line for transmitting data or instructions, and plays a vital role in the urban rail transit system. Furthermore, "communication signal cable" refers to a physical medium that connects different communication nodes and is responsible for transmitting information (such as control commands, monitoring data, etc.). When the line is in good health, these cables can ensure efficient communication between the train control system and the ground station and maintain normal driving order. However, once damaged or falling off, it may cause serious safety hazards and even cause accidents. Therefore, regularly checking and maintaining the status of communication signal cables to prevent them from failing due to aging or other reasons has become an indispensable part of ensuring the safe operation of urban rail transit.

[0012] In the present invention, "detachment detection" is a mechanism designed to prevent potential dangers, which is used to detect whether the communication signal cable has detached from its original position. Furthermore, "detachment detection" is a technical means specifically used to identify whether the communication signal cable has been displaced or fallen from a predetermined position. This work is very important because any unexpected movement of the cable may interfere with the normal communication link and thus affect the safe operation of the train. Traditional detection methods often rely on manual inspections, but this method is inefficient and prone to missing hidden dangers. In contrast, the detection method and system provided by the present invention can locate problem points more quickly and accurately, and issue alarms in a timely manner, thereby greatly reducing the risk of accidents.

[0013] The present invention discloses a method for detecting the cable shedding of a two-stage tunnel surface communication signal, which comprises the following steps:

[0014] S1. Using the image acquisition module to dynamically acquire original images of tunnel surface auxiliary facilities and equipment;

[0015] S2, preprocessing the original image acquired by the image acquisition module;

[0016] S3. Classify the tunnel's apparent auxiliary facilities and equipment based on the pre-processed image data;

[0017] S4. Detect the slope of the communication signal cable to detect whether the communication signal cable is detached.

[0018] The detection method first dynamically collects the original images of the tunnel's apparent auxiliary facilities and equipment, which not only captures the real-time information of the changes in the tunnel's internal environment, but also avoids the problem of possible omissions in the traditional manual inspection method due to the long time interval. Next, by preprocessing the collected original images, the image quality and feature recognizability can be significantly improved, thus laying a solid foundation for subsequent classification and detection work. Finally, on the basis of classifying the preprocessed image data, the slope of the communication signal cable is detected to achieve effective identification of the communication signal cable falling off. This process not only improves the accuracy of detection, but also greatly reduces the possibility of false alarms, ensuring the safe operation of the urban rail transit system. The entire detection process is completed in an automated manner, which not only reduces labor costs, but also improves work efficiency, making the detection work more efficient and timely. In addition, this method can adapt to different types of tunnel environments, and can complete the detection during normal train operation, further improving the safety and reliability of the detection.

[0019] According to a preferred embodiment, the image acquisition module is mounted on a carrier vehicle so that when the carrier vehicle moves on the track, the image acquisition module can respond to the received pulse signal to implement acquisition triggering based on the preset second basic information including the trigger mode and trigger parameters, wherein the trigger mode includes time triggering and distance triggering.

[0020] The image acquisition module is mounted on the carrier vehicle, and the acquisition trigger is realized by responding to the pulse signal according to the preset trigger mode (such as time trigger or distance trigger). This design cleverly combines the characteristics of train operation with the detection requirements. In practical applications, when the carrier vehicle moves along the track, the image acquisition module installed on it can automatically start the image acquisition process according to the set trigger parameters. For the distance trigger mode, the accurate calculation of the encoder frequency and the wheel circumference ensures that each acquisition occurs at the correct position, which not only improves the spatial consistency of the image data, but also provides a reliable data basis for subsequent analysis. In addition, by adjusting the trigger mode and parameter settings, the acquisition frequency can be flexibly controlled according to the specific conditions of different sections, which not only ensures the comprehensiveness of the detection, but also avoids unnecessary repeated operations, thereby effectively improving the detection efficiency. Such a design scheme not only meets the requirements of efficient operation of modern rail transit, but also provides technical support for the realization of intelligent and automated detection, especially under the condition of high-speed operation, it can still maintain high-precision image acquisition, greatly enhancing the applicability and robustness of the detection system.

[0021] According to a preferred embodiment, the preprocessing process of the original image acquired by the image acquisition module includes image geometry correction using affine transformation, noise removal using Gaussian filtering, and image contrast enhancement using Retinex algorithm to improve image quality and feature recognizability.

[0022] In the process of preprocessing the original image acquired by the image acquisition module, affine transformation is used for image geometric correction, Gaussian filtering is used for noise removal, and Retinex algorithm is used for image contrast enhancement. These steps successively solve the problems of geometric distortion, random noise interference, and uneven illumination in the image. First, affine transformation eliminates deformation caused by shooting angle or position deviation by adjusting parameters such as image angle and scaling, ensuring that all images have a consistent standard view. Then, Gaussian filtering effectively smoothes the high-frequency noise in the image while retaining edge details, making the image clearer and cleaner. Finally, the Retinex algorithm enhances the contrast of local areas by separating the reflection component and illumination component in the image, so that the originally dim and unclear details can be revealed. After the above series of preprocessing operations, the image quality has been significantly improved and the features are more prominent, which is not only conducive to improving the accuracy of subsequent classification and recognition, but also provides high-quality data support for the final communication signal cable detachment detection. In addition, these preprocessing steps can also reduce the computational burden in subsequent processing and improve processing speed and response time.

[0023] According to a preferred embodiment, the YOLO v5m algorithm is used to classify the tunnel's apparent ancillary facilities and equipment for the preprocessed image data, so as to identify different types of equipment and store the tunnel's apparent ancillary facilities and equipment images in corresponding databases according to the types of equipment contained therein, wherein YOLO v5m is used as the basic model, and a SE module including a compression process and an excitation process is introduced in the backbone layer of YOLO v5m.

[0024] Based on the preprocessed image data, the YOLO v5m algorithm is used to classify the tunnel's apparent auxiliary facilities and equipment, and the images of different types of equipment identified are stored in the corresponding databases. This approach fully utilizes the advantages of deep learning technology. As an advanced target detection algorithm, YOLO v5m is fast and accurate, and can process a large number of images in a short time and give reliable classification results. In particular, after the SE module is introduced into its backbone layer, the model can adaptively recalibrate the channel feature response, further enhance useful features, and suppress irrelevant information. This means that in the face of complex and changing tunnel environments, the YOLO v5m+SE model can more accurately identify various types of communication signal cables and related equipment, reducing the possibility of misjudgment. In addition, the practice of storing the recognition results by category not only facilitates subsequent data management and query, but also provides convenience for long-term monitoring and trend analysis. In this way, the system not only improves the efficiency and accuracy of detection, but also provides detailed reference materials for maintenance personnel, which helps to formulate scientific and reasonable maintenance plans and ensure the safe and reliable operation of urban rail transit systems. This classification method can also continuously improve its performance and accuracy through machine learning self-optimization as new data continues to accumulate.

[0025] According to a preferred embodiment, the image of each communication signal cable in the communication signal cable library is used as an input image, edge detection is performed on the input image and Hough transform is applied, voting is accumulated in the parameter space to identify the straight line segment of the communication signal cable, and then the specific position of the cable is reconstructed according to the local maximum value in the accumulator array, so as to determine the slope of the communication signal cable in the input image.

[0026] For each communication signal cable image in the communication signal cable library, through the application of edge detection and Hough transform, votes can be accumulated in the parameter space to identify the straight line segment of the cable, and the specific position of the cable can be reconstructed according to the local maximum value in the accumulator array, so as to determine the slope of the communication signal cable in the input image. This method cleverly uses the principle of mathematical transformation to simplify complex image processing problems. First, edge detection can extract pixels that potentially belong to a straight line from the image, providing basic data for the subsequent Hough transform. Then, the Hough transform realizes efficient search for straight line features by mapping each edge point to ρ (the distance from the straight line to the origin) and θ (the angle between the straight line and the positive direction of the x-axis) in the parameter space. In this process, the accumulator array plays a key role: it records the number of votes for each possible straight line, so that those straight line segments that actually exist in the image can be highlighted. Finally, by analyzing the local maximum value in the accumulator array, the position and slope of the communication signal cable can be accurately determined. This method is not only unaffected by the rotation and scaling of the straight line, but also can effectively deal with the noise and occlusion problems in the image, greatly improving the robustness and accuracy of detection. Through this precise slope detection, abnormal conditions of communication signal cables can be discovered in a timely manner, and potential safety hazards can be warned in advance, providing strong technical support for ensuring the safety of urban rail transit.

[0027] According to a preferred embodiment, a cable slope threshold is set to detect the detachment of a communication signal cable by combining the cable slope threshold and the abnormal characteristics of a cable bracket, wherein if the slope of any communication signal cable exceeds the cable slope threshold, it is judged as suspected detachment, and then whether the communication signal cable has detached is determined by judging whether an abnormal condition including position displacement, posture change and / or structural loss has occurred in a cable bracket associated with the suspected detached communication signal cable.

[0028] The above scheme provides an intelligent judgment mechanism based on multi-feature fusion. First, through the pre-set cable slope threshold, the system can quickly screen out suspected detached cables whose absolute slope value exceeds the cable slope threshold (for example, 0.3). This threshold is an empirical value based on a large amount of data analysis and has a high degree of credibility. Subsequently, for these suspected detached cables, the cable brackets associated with them are further checked for abnormal conditions such as position offset, posture change, or structural loss. This two-step approach not only takes into account the physical properties of the cable itself, but also pays attention to the status of its fixing device, forming a more comprehensive evaluation system. Specifically, if the cable slope exceeds the threshold and the corresponding bracket is abnormal, it can be more confidently determined that the cable has fallen off. On the contrary, if only one of the conditions is met, further investigation and confirmation is required. This method is not only simple and easy to implement, does not require additional complex algorithm support, but also has a high accuracy rate. It can significantly improve the sensitivity and reliability of communication signal cable detachment detection without affecting the overall performance of the system. Through this double verification mechanism, not only can the accuracy of detection be improved, but also the situation of false alarms can be reduced, providing maintenance personnel with a more reliable decision-making basis and ensuring the safe operation of urban rail transit.

[0029] The present invention also discloses a two-stage tunnel apparent communication signal cable detachment detection system, which includes: an image acquisition module, used to dynamically acquire original images of tunnel apparent ancillary facilities and equipment; a processing module, used to pre-process the original images obtained by the image acquisition module, and then classify the tunnel apparent ancillary facilities and equipment based on the pre-processed image data, and detect the slope of the communication signal cable to realize the detachment detection of the communication signal cable.

[0030] The present invention constructs a two-stage tunnel apparent communication signal cable detachment detection system, which integrates functional units such as image acquisition modules and processing modules, and aims to dynamically collect original images of tunnel apparent auxiliary facilities and equipment, and perform a series of operations such as pre-processing, classification, and slope detection, and finally achieve effective monitoring of communication signal cable detachment. Compared with the traditional manual inspection method, this system not only greatly improves the automation level of detection work, reduces the interference of human factors, but also ensures the consistency and accuracy of data collection. More importantly, through integrated hardware design and software algorithm optimization, the system can complete a large number of image processing and analysis tasks in a short time, greatly improving the detection efficiency.

[0031] According to a preferred embodiment, it is configured with an integrated device externally placed on a carrier vehicle and a mounting bracket for connecting the integrated device to the carrier vehicle, and the image acquisition module is installed inside the integrated device, wherein the mounting bracket includes two non-coplanar mounting surfaces, the first mounting surface is used to connect to the carrier vehicle, and the second mounting surface is used to connect to the integrated device.

[0032] The detection system is equipped with an integrated device that is externally mounted on the carrier vehicle and fixed with a mounting bracket with two non-coplanar mounting surfaces. This design not only simplifies the installation process of the equipment, but also enhances the stability and applicability of the system. The compact structure of the integrated device makes it easy to install and disassemble, while ensuring the close fit between the various components and reducing the impact of the external environment on its working performance. The first mounting surface of the mounting bracket is connected to the carrier vehicle, and the second mounting surface is connected to the integrated device. This non-coplanar design cleverly solves the spatial layout problem and ensures the versatility and compatibility of the integrated device on various vehicle models. In addition, the stable connection method of the mounting bracket provides a guarantee for the stability of the equipment during high-speed driving, preventing loosening or falling off due to vibration or impact. This design not only facilitates the operation of on-site construction personnel, but also extends the service life of the equipment and reduces maintenance costs.

[0033] According to a preferred embodiment, the integrated device is configured with a shell portion including an internal cavity, and a mounting groove for connecting to the second mounting surface of the mounting bracket is provided on the bottom plate of the shell portion, so that the integrated device can be detachably fixed on the mounting bracket.

[0034] A mounting groove for connecting to the second mounting surface of the mounting bracket is provided on the bottom plate of the outer shell, so that the integrated device can be detachably fixed on the mounting bracket. This design reflects the advantages of the modular design concept. First, the existence of the mounting groove makes the connection between the outer shell and the mounting bracket simpler and faster, which greatly facilitates on-site commissioning and maintenance work. Secondly, the detachable fixing method not only allows the user to flexibly adjust the position and angle of the equipment according to actual needs, but also facilitates the inspection or upgrading of internal components without worrying about damaging the overall structure. In addition, this modular design also helps to improve the scalability and flexibility of the system. For example, new functional modules can be added or existing components can be replaced in the future to adapt to changing application scenarios and technological advances.

[0035] According to a preferred embodiment, the shell of the integrated device includes two oppositely arranged side panels and a working plate arranged between the two side panels, and a plurality of through holes are provided on the working plate at intervals so that the linear array camera and the laser fill light of the image acquisition module can be installed on these through holes in a set manner, wherein the opening position of the through holes on the working plate can be determined based on the field of view angle of the linear array camera, so as to achieve imaging of the tunnel lining surface excluding the track area while ensuring that the field of view ranges of adjacent linear array cameras overlap at least partially.

[0036] This design takes into full account the actual needs and optical characteristics of image acquisition. First, the position of the through-holes is carefully arranged based on the field of view of the linear array camera to ensure that the field of view of adjacent linear array cameras overlaps at least partially, thereby achieving seamless coverage of the tunnel lining surface except for the track area. This design not only improves the integrity of image acquisition, but also avoids information loss caused by blind spots in the field of view. Secondly, the complete installation of the linear array camera and the laser fill light ensures the coordination between the two, that is, sufficient lighting is provided while shooting, ensuring that the image quality is not limited by light conditions. In addition, by rationally planning the position and spacing of the through-holes, the overall layout of the equipment can be optimized, the volume occupied can be reduced, and the space utilization rate can be improved. Finally, such a design enables the equipment to adapt to different types of tunnel environments, whether it is a tunnel section of different widths or a complex internal structure, it can obtain high-quality image data, providing a solid foundation for subsequent analysis and processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 is a flow chart of the steps of the detection method provided by the present invention;

[0038] Figure 2 This is a schematic diagram of setting the second basic information of an image acquisition module in a preferred embodiment provided by the present invention;

[0039] Figure 3 It is a schematic diagram of the collected imaging of different types of tunnel surface auxiliary facilities and equipment provided by the present invention;

[0040] Figure 4 It is a schematic diagram of the communication signal cable shedding detection process provided by the present invention;

[0041] Figure 5 It is a hardware connection diagram of the detection system provided by the present invention;

[0042] Figure 6 It is a schematic diagram of the structure of the mounting bracket of the detection system provided by the present invention;

[0043] Figure 7 It is a structural schematic diagram of an integrated device of a detection system provided by the present invention;

[0044] Figure 8 It is a schematic diagram of the imaging range of the image acquisition module provided by the present invention.

[0045] Reference numerals list

[0046] 100: carrier vehicle; 200: mounting bracket; 211: first mounting surface; 212: second mounting surface; 300: integrated device; 310: outer shell; 320: bottom plate; 321: mounting groove; 330: side plate; 340: working plate; 341: through hole; 342: image acquisition module; 343: linear array camera; 344: field of view; 345: laser fill light; 350: handle; 400: processing module; 500: tunnel lining; 600: auxiliary facilities and equipment; 610: junction box; 620: broadcast speaker; 630: communication signal cable; 640: cable bracket; 650: instrument; 660: signal machine; 670: signal machine box; 680: socket box; 690: CCTV equipment system. DETAILED DESCRIPTION

[0047] The following is a detailed description with reference to the accompanying drawings.

[0048] Example 1

[0049] like Figure 1 As shown, the present invention discloses a method for detecting the detachment of a two-stage tunnel surface communication signal cable 630, which may include the following steps:

[0050] S1. Using the image acquisition module 342 to dynamically acquire the original image of the tunnel's apparent auxiliary facilities and equipment 600;

[0051] S2, preprocessing the original image acquired by the image acquisition module 342;

[0052] S3, classifying the tunnel appearance auxiliary facilities and equipment 600 based on the pre-processed image data;

[0053] S4. Detect the slope of the communication signal cable 630 to detect whether the communication signal cable 630 is detached.

[0054] Preferably, in step S1, the image acquisition module 342 installed on the train (e.g., electric passenger car) can be used to realize digital imaging of the surface of the tunnel lining 500 during the operation of the vehicle, wherein even if the train runs at a higher speed (e.g., 80 km / h or 120 km / h), the image acquisition module 342 can still acquire images of the tunnel's apparent auxiliary facilities and equipment 600, thereby overcoming the problems of low efficiency, high cost, and poor detection safety in methods such as manual inspection or using an electric drive integrated detection platform equipped with a detection system to detect at a speed of 15 km / h. In order to ensure the quality and reliability of the image data, the image acquisition module 342 can select a suitable camera, wherein the image acquisition module 342 can select an industrial-grade camera with high resolution, low noise, and a wide dynamic range, and customize the adjustment according to the characteristics of the tunnel environment (e.g., light conditions, humidity, temperature, etc.). These cameras can adapt to the relatively dim environment inside the tunnel and can maintain the stability and clarity of the image during high-speed movement. Furthermore, considering the camera's viewing angle coverage and installation stability, the image acquisition module 342 can be installed at the front or rear of the train to ensure that its field of view can cover the entire tunnel's apparent auxiliary facilities and equipment 600. In addition, it is also necessary to consider the camera's protective measures, such as waterproof, dustproof and earthquake-resistant design, to cope with the complex environment in the tunnel.

[0055] Preferably, due to the poor lighting conditions inside the tunnel, the fill light of the image acquisition module 342 is particularly important. The fill light may include an LED light bar or other types of light sources, which can provide uniform and sufficient lighting, so that the camera can obtain high-quality images even in low light conditions. The brightness and angle of the fill light need to be adjusted according to the actual situation to avoid problems such as overexposure or shadows that affect the image quality.

[0056] Preferably, after the image acquisition module 342 is installed, the first basic information such as the exposure parameter and gain parameter of the camera and the fill light parameter of the fill light can be adjusted, and the second basic information such as the trigger mode and trigger parameter can also be set, so that when the train (hereinafter referred to as the carrier vehicle 100) equipped with the image acquisition module 342 moves on the track, the image acquisition module 342 can receive a pulse signal to realize the acquisition trigger. Further, the trigger mode determines the rule according to which the image acquisition operation is started. Different trigger modes can be associated with a corresponding trigger parameter. Common trigger modes include time trigger and distance trigger. In addition, other triggering methods may also include event triggering. The trigger parameter is the trigger condition data set for the selected trigger mode. When the trigger parameter corresponding to the trigger mode reaches the set trigger condition data, the rule for starting the image acquisition operation is met, and the image acquisition module 342 can be driven to perform image acquisition.

[0057] For example, Figure 2 As shown, for the time-triggered trigger mode, a preset trigger time (for example, 300ms) can be input, so that the image acquisition module 342 can be driven to perform image acquisition every time the preset trigger time interval occurs; for the distance-triggered trigger mode, in addition to inputting a preset trigger distance (for example, 750mm), the encoder frequency (that is, the number of pulses output per revolution) and the wheel circumference can also be input to calculate the distance traveled by the carrier vehicle 100, thereby accurately determining whether the carrier vehicle 100 has reached the preset trigger distance.

[0058] Preferably, based on the set first basic information and second basic information, the image acquisition module 342 can collect original images of the tunnel's apparent ancillary facilities and equipment 600 at intervals during the movement of the carrier vehicle 100, wherein the ancillary facilities and equipment 600 may include a junction box 610, a broadcast speaker 620, a communication signal cable 630, a cable bracket 640, an instrument 650, a signal machine 660, a signal machine box 670, a socket box 680, a CCTV equipment system 690, and the like.

[0059] Preferably, in step S2, the original image acquired by the image acquisition module 342 is often affected by various factors, such as uneven illumination, motion blur, noise, etc., which will reduce the accuracy of subsequent analysis. Therefore, a series of preprocessing operations must be performed on these image data to improve the image quality and the recognizability of features. Preferably, the preprocessing process may include image correction, noise removal, image contrast enhancement, etc., to improve the image quality and the recognizability of features.

[0060] In a tunnel environment, due to the speed change of the carrier vehicle 100 and the different curve radius of the track, the collected original image may have problems such as tilt, stretching or compression. In addition, the angle setting of the image acquisition module 342 may also cause image deformation. In order to ensure that all images can accurately reflect the actual scene and provide a consistent basis for subsequent feature extraction, it is necessary to first perform geometric correction of the original image. Preferably, in step S2, affine transformation can be used for image geometric correction, wherein affine transformation is a linear mapping method, which can be expressed in the form of matrix multiplication and is suitable for describing geometric transformations such as translation, rotation, scaling and shearing. By calculating the correspondence between specific points, appropriate transformation parameters can be determined to convert the original image into a standard view that meets the expectations. Preferably, algorithms such as SIFT (Scale-Invariant Feature Transform) or ORB (Oriented FAST and Rotated BRIEF) can be used to find representative feature points in the original image and match them with corresponding points in the standard template. Based on the above matching results, the least squares method or other optimization algorithms are used to solve the optimal affine transformation matrix. This matrix will be used to adjust the position, direction and size of the entire image to make it as close to the ideal state as possible. Finally, the calculated transformation matrix is ​​applied to the original image to complete the geometric correction. At the same time, the quality of the corrected image needs to be evaluated to check whether there is distortion, and the parameters are adjusted according to the actual situation until they are satisfactory.

[0061] Despite the use of high-quality cameras and fillers, the images collected during high-speed movement still inevitably contain certain noise components, such as randomly distributed bright spots or dark spots. These noises not only affect the visual effect, but also interfere with subsequent feature detection and classification. Therefore, it is necessary to take effective noise reduction measures to purify the image data. Preferably, in step S2, Gaussian filtering can be used for noise removal, wherein Gaussian filtering is a spatial domain filter that smoothes the image by weighted averaging of neighboring pixels, where the weight is determined by a two-dimensional Gaussian function. This method can effectively reduce high-frequency noise while retaining edge information, and is particularly suitable for processing images in natural scenes. The performance of the filter depends on the selection of two main parameters: its core size and standard deviation. A larger core size can eliminate noise more thoroughly, but may also cause image blur; a smaller core size may not be able to fully filter noise. Therefore, it is necessary to flexibly adjust these two parameters according to the specific application scenario to achieve the best balance. Considering that different types of noise are distributed on different spatial scales, a multi-scale Gaussian filtering strategy can be adopted, that is, a filter with a larger core size is used to process low-frequency noise first, and then the core size is gradually reduced to remove high-frequency noise, and finally a clean and clear image is obtained. In order to avoid artifacts at the edges of the image, mirror reflection, period extension or constant padding can be selected to ensure the consistency and integrity of the filtering operation.

[0062] The lighting conditions in the tunnel are complex and changeable. Some areas may have strong light and shadow contrast, while other places appear dim and unclear. This uneven light distribution makes it difficult to use the image directly for analysis, especially when the target objects are similar in color. To this end, it is necessary to enhance the contrast of the image to make each detail more prominent and facilitate subsequent feature recognition. Preferably, in step S2, the Retinex algorithm can be used to enhance the image contrast, wherein the Retinex algorithm can separate the reflection component (i.e., the true color of the object) and the illumination component in the image, and then recombine them to improve the overall contrast. The Retinex algorithm may include single-scale Retinex (abbreviated as SSR) and multi-scale Retinex (abbreviated as MSR). Considering the diversity of lighting conditions in the tunnel, an adaptive parameter adjustment strategy may be adopted. For example, the weights of each scale are dynamically adjusted according to the brightness level of different areas, or the appropriate algorithm version is automatically selected according to the image content to ensure the best enhancement effect. After completing the Retinex process, histogram equalization, gamma correction and other technologies can be further applied to fine-tune the overall tone of the image so that it not only maintains the original style characteristics, but also has good visual effects.

[0063] Preferably, in step S3, the YOLO v5m algorithm can be used to classify the tunnel appearance auxiliary facilities and equipment 600 to identify the equipment such as the junction box 610, the broadcast speaker 620, the communication signal cable 630, the cable bracket 640, the meter 650, the signal machine 660, the signal machine box 670, the socket box 680, the CCTV equipment system 690, etc. Figure 3 Furthermore, during the training phase, the collected images of various tunnel appearance auxiliary facilities and equipment 600 can be set with corresponding labels to serve as the training set of the model; during the testing phase, the model will automatically add corresponding labels after identifying the equipment.

[0064] Preferably, based on comprehensive considerations of detection accuracy, model size, and detection speed, the present invention selects YOLO v5m as the basic model, and introduces the SE module in the backbone layer of YOLO v5m. The SE module includes two processes: compression and excitation.

[0065] Preferably, the compression process compresses the spatial dimension of the feature map from H×W×C to 1×1×C through global average pooling, so that the features of each channel are aggregated into a global statistic. In this way, the statistics of each channel contain global spatial information, which is equivalent to obtaining the receptive field of the entire image. This embedding of global information helps the model capture the global contextual information in the image, thereby improving the accuracy of image recognition. The specific formula is as follows:

[0066]

[0067] Among them, u C (i, j) represents the element in the i-th row and j-th column of the C-th channel in the input feature map U. H and W are the height and width of the feature map, respectively, and C represents the number of channels. After the compression operation, the feature map is compressed into a 1×1×C vector, and each channel corresponds to a global average value. C Represents the compressed feature vector.

[0068] Preferably, the excitation process predicts and adaptively recalibrates the importance of each channel through two fully connected layers. Furthermore, the first fully connected layer reduces the dimension of the 1×1×C feature vector to 1×1×C×Ratio, and the second fully connected layer increases it to 1×1×C, where Ratio is a scaling parameter used to reduce the number of channels and reduce the amount of computation. Through the excitation process, the model can learn the dependencies between different channels and assign a weight to each channel, which represents the importance of the channel to the final task.

[0069] Preferably, the excitation process may include two fully connected layers and a nonlinear activation function for learning to generate a weight vector of a channel. The specific formula is as follows:

[0070] y C =σ(W 2 ·δ(W 1 ·z C )),

[0071] Among them, z C is the compressed feature vector, W 1 and W 2 are the weights of the two fully connected layers, δ is the ReLU activation function, σ is the Sigmoid activation function, and y C is the output of the excitation operation, i.e., the channel weight vector.

[0072] The channel weight vector can be used to weight each channel of the original feature map and complete the recalibration of the original features in the channel dimension. This recalibration helps the model focus on more informative features and suppress less important features, thereby improving image recognition performance. The specific formula is as follows:

[0073]

[0074] in, Represents the weighted feature map, y C Represents the channel weight vector of the Cth channel. The new feature map will be sent to the next layer of the network for subsequent feature extraction and image recognition tasks.

[0075] Through the above steps, the detection accuracy can be improved, and the identification of tunnel surface ancillary facilities and equipment 600 such as the distribution box 610, broadcast speaker 620, communication signal cable 630, cable bracket 640, instrument 650, signal machine 660, signal machine box 670, socket box 680, CCTV equipment system 690 can be achieved.

[0076] For example, assuming that the size of the feature map X is 2×2×3, it is a three-dimensional tensor, where the first dimension is the height (ie, H=2), the second dimension is the width (ie, W=2), and the third dimension is the number of channels (ie, C=3), as shown below:

[0077]

[0078] Here, X consists of three 2×2 matrices, each of which represents the features of a channel.

[0079] After the compression process of the SE module, the feature map of each channel is globally averaged to obtain a 1×1×3 feature vector z:

[0080]

[0081] Next, the importance weight of each channel is learned through the excitation process. Assume that the channel weight vector y obtained after the fully connected layer and activation function is as follows:

[0082]

[0083] Now, the channel weight vector obtained by the excitation operation is multiplied by each channel of the original feature map channel by channel to obtain the weighted feature map X':

[0084]

[0085] From this, we can see that for the weighted feature map X', the value of each channel is adjusted according to its importance. This new feature map will be used in subsequent layers of the network to help the model better identify features in the image.

[0086] Preferably, the images of various types of tunnel surface ancillary facilities and equipment 600 identified in step S3 can be stored in corresponding databases respectively. For example, the images of communication signal cables 630 can be stored in a communication signal cable library, so that in step S4, each communication signal cable 630 in the communication signal cable library can be detected using the Hough line detection method.

[0087] Preferably, in step S4, edge detection can be performed on the input image first to determine the pixel points that may belong to the straight line, and define the parameter space, wherein edge detection can use algorithms such as Canny and Sobel. Exemplarily, the use of the Canny algorithm can obtain a good signal-to-noise ratio and sub-pixel accuracy. Preferably, in the Hough transform, a straight line can be defined by two parameters: ρ (the distance from the origin to the straight line) and θ (the angle between the straight line and the positive direction of the x-axis). Therefore, for each edge point in the image, all possible combinations of ρ and θ can be calculated, and these combinations define the position of the straight line in the parameter space. Further, an accumulator array is created, which is a two-dimensional array, whose horizontal coordinate corresponds to θ, the vertical coordinate corresponds to ρ, and each unit corresponds to a point (ρ, θ) in the parameter space, which is used to store the count of each possible straight line. If multiple edge points correspond to the same combination of ρ and θ, then the count of this combination will increase, indicating that multiple edges may belong to the same straight line. By finding the local maximum in the accumulator array, the ρ and θ combinations corresponding to these maximum values ​​represent the detected straight line, and then returning the corresponding ρ and θ values ​​according to the position of the local maximum in the accumulator array, it can be used to reconstruct the straight line in the original image, that is, the specific position of the cable. For example, the conversion from parameter space back to image space can be achieved by using the conversion relationship between polar coordinate system and Cartesian coordinate system.

[0088] When the above method is used to detect straight lines in an image, it is not affected by the rotation and scaling of the straight line, that is, no matter how the straight line is tilted in the image or how its length changes, straight lines close to horizontal, vertical or diagonal can be accurately detected. The Hough line detection method can handle the noise in the image well. Even if there are noise points after edge detection, the image contrast is low, or even the image is partially blocked or broken, the entire straight line can be detected. In addition, the Hough transform can control the sensitivity of the detection by adjusting the accumulator threshold, thereby making a trade-off between the number of detected straight lines and accuracy.

[0089] Preferably, in step S4, a cable slope threshold may be set to detect the fall-off of the communication signal cable 630 by combining the cable slope threshold and the abnormality of the cable bracket 640. Figure 4 As shown. Preferably, in the present invention, the cable slope threshold can be set to 0.3. This is because by analyzing a large amount of image data, it is found that when the communication signal cable 630 falls off, the absolute value of its slope is above 0.3. Therefore, if the absolute value of the cable slope is greater than 0.3, it is judged as suspected falling off. Preferably, the cable bracket 640 is usually used to fix the communication signal cable 630, and these cable brackets 640 should maintain a certain position and posture, that is, the position of the cable bracket 640 relative to the tunnel wall or track is fixed. Further, for the cable bracket 640 associated with the suspected fallen communication signal cable 630, the image of the corresponding cable bracket 640 is extracted from the cable bracket library generated in step S3, so as to compare with the standard image of the cable bracket 640 to determine whether the cable bracket 640 associated with the suspected fallen communication signal cable 630 has abnormal conditions such as position displacement, posture change and / or structural loss, wherein, if the above abnormal conditions exist, it can be determined that the communication signal cable 630 falls off. This method is simple and easy, does not require the use of additional algorithms, and has high accuracy.

[0090] Example 2

[0091] This embodiment is a further improvement on Embodiment 1, and the repeated contents will not be repeated here.

[0092] The present invention discloses a two-stage tunnel apparent communication signal cable 630 detachment detection system, which can execute the detection method as described in Example 1 to realize intelligent identification of tunnel apparent communication signal cable 630 detachment defects.

[0093] Preferably, if Figure 5As shown, the detection system of the present invention may include: an image acquisition module 342, which is used to dynamically acquire the original image of the tunnel's apparent ancillary facilities and equipment 600; a processing module 400, which is used to pre-process the original image acquired by the image acquisition module 342, and then classify the tunnel's apparent ancillary facilities and equipment 600 based on the pre-processed image data, and detect the slope of the communication signal cable 630 to achieve the detachment detection of the communication signal cable 630.

[0094] Preferably, if Figure 6 and Figure 7 As shown, the detection system of the present invention can be configured with an integrated device 300 externally mounted on a carrier vehicle 100 and a mounting bracket 200 for connecting the integrated device 300 to the carrier vehicle 100. The integrated device 300 can be connected to the front or rear of the carrier vehicle 100 through the mounting bracket 200, wherein the mounting bracket 200 can be configured as a Z-shaped structure to form two parallel but non-coplanar mounting surfaces. Further, the first mounting surface 211 with a relatively larger area of ​​the two mounting surfaces is used to connect to the body of the carrier vehicle 100, and the second mounting surface 212 with a relatively smaller area is used to connect to the integrated device 300.

[0095] Preferably, the image acquisition module 342 can be arranged in the integrated device 300, and a synchronous controller for generating control signals and an industrial computer for realizing data processing can also be arranged in the integrated device 300. Preferably, the synchronous controller can be used to adjust the first basic information and / or the second basic information, and can generate a trigger signal when it is judged that the trigger condition is met, so that the image acquisition module 342 completes the acquisition of the original image. Preferably, the processing module 400 can be integrated in the industrial computer to realize synchronous completion of data acquisition and processing in the integrated device 300, and can also be arranged in an external server host that is communicatively connected to the industrial computer, and the server host can be, for example, arranged on the carrier vehicle 100. Furthermore, a power supply host can also be arranged on the carrier vehicle 100 to provide a stable power supply for the operation of the integrated device 300.

[0096] Preferably, if Figure 7As shown, the integrated device 300 may have a housing portion 310, wherein a mounting groove 321 for connecting with the second mounting surface 212 is provided on the bottom plate 320 of the housing portion 310, so that the integrated device 300 can be detachably fixed on the mounting bracket 200. Further, the housing portion 310 may include two oppositely disposed side plates 330 and a working plate 340 disposed between the two side plates 330, wherein the two sides of the working plate 340 are respectively fitted with the edges of the two side plates 330, and the two ends of the working plate 340 are fitted with the edges of the bottom plate 320, so as to form a closed structure with an internal cavity. Preferably, a plurality of through holes 341 may be provided on the working plate 340, so that the camera and the fill light of the image acquisition module 342 can realize image acquisition and fill light through these through holes 341. Preferably, the camera of the image acquisition module 342 can be configured as a (high-definition) line array camera 343, the fill light can be configured as a laser fill light 345, and the line array camera 343 and the laser fill light 345 can be combined in a set, so that each through hole 341 on the working plate 340 can be provided with a set of line array cameras 343 and laser fill light 345, so as to achieve coordinated use. Preferably, the integrated device 300 of the present invention can be provided with eight sets of line array cameras 343 and laser fill light 345, that is, including eight line array cameras 343 and eight laser fill light 345, and these eight sets of line array cameras 343 and laser fill light 345 can be arranged on the working plate 340 at intervals in different orientations, wherein, for example Figure 8 As shown, the field of view angle of each line array camera 343 is 45°, and the overlapping angle of the field of view range 344 of adjacent line array cameras 343 is approximately 10°, so that the eight line array cameras 343 can achieve 290° imaging of the tunnel (lining surface), that is, excluding the track area within 70° below the tunnel, so as to cover most of the tunnel's apparent ancillary facilities and equipment 600.

[0097] Preferably, if Figure 7 As shown, a handle 350 for users to carry or grab can also be provided on the working plate 340 of the integrated device 300, wherein the handle 350 can be arranged in a manner that does not block the optical path of the linear array camera 343 and the laser fill light 345.

[0098] It should be noted that the above-mentioned specific embodiments are exemplary, and those skilled in the art can come up with various solutions inspired by the disclosure of the present invention, and these solutions also belong to the disclosure scope of the present invention and fall within the protection scope of the present invention. Those skilled in the art should understand that the present invention specification and its drawings are illustrative and do not constitute a limitation of the claims. The scope of protection of the present invention is defined by the claims and their equivalents. The present invention specification contains multiple inventive concepts, such as "preferably" or "according to a preferred embodiment", which means that the corresponding paragraph discloses an independent concept, and the applicant reserves the right to file a divisional application based on each inventive concept. Throughout the text, the features guided by "preferably" are only an optional method and should not be understood as a must-have setting. Therefore, the applicant reserves the right to abandon or delete the relevant preferred features at any time.

Claims

1. A communication signal cable detachment detection method, characterized in that: It includes the following steps: S1, using an image acquisition module (342) to dynamically acquire original images of tunnel surface auxiliary facilities and equipment (600); S2, preprocessing the original image acquired by the image acquisition module (342); S3, classifying the tunnel appearance auxiliary facilities and equipment (600) based on the pre-processed image data; S4. Detect the slope of the communication signal cable (630) to detect whether the communication signal cable (630) has fallen off.

2. The detection method according to claim 1, characterized in that: The image acquisition module (342) is mounted on a carrier vehicle (100) so that when the carrier vehicle (100) moves on a track, the image acquisition module (342) can respond to a received pulse signal to implement acquisition triggering based on preset second basic information including a trigger mode and trigger parameters, wherein the trigger mode includes a time trigger and a distance trigger.

3. The detection method according to claim 1 or 2, characterized in that: The preprocessing process of the original image acquired by the image acquisition module (342) includes using affine transformation to perform image geometric correction, using Gaussian filtering to remove noise, and using Retinex algorithm to enhance image contrast, so as to improve image quality and feature recognizability.

4. The detection method according to any one of claims 1 to 3, characterized in that: For the preprocessed image data, the YOLO v5m algorithm is used to classify the tunnel apparent ancillary facilities and equipment (600) to identify different types of equipment and store the tunnel apparent ancillary facilities and equipment (600) images in corresponding databases according to the types of equipment contained therein, wherein YOLO v5m is used as the basic model, and the SE module including the compression process and the excitation process is introduced into the backbone layer of YOLO v5m.

5. The detection method according to any one of claims 1 to 4, characterized in that: The image of each communication signal cable (630) in the communication signal cable library is used as an input image. By performing edge detection on the input image and applying Hough transform, voting is accumulated in the parameter space to identify the straight line segment of the communication signal cable (630), and then the specific position of the cable is reconstructed according to the local maximum value in the accumulator array, so as to determine the slope of the communication signal cable (630) in the input image.

6. The detection method according to any one of claims 1 to 5, characterized in that: A cable slope threshold is set to detect the detachment of the communication signal cable (630) by combining the cable slope threshold and the abnormal characteristics of the cable bracket (640), wherein if the slope of any communication signal cable (630) exceeds the cable slope threshold, it is judged as suspected detachment, and then it is determined whether the communication signal cable (630) has detached by judging whether the cable bracket (640) associated with the suspected detached communication signal cable (630) has abnormal conditions including position displacement, posture change and / or structural loss.

7. A communication signal cable detachment detection system, characterized in that: It includes: An image acquisition module (342) for dynamically acquiring original images of tunnel surface auxiliary facilities and equipment (600); The processing module (400) is used to pre-process the original image acquired by the image acquisition module (342), and then classify the tunnel surface auxiliary facilities and equipment (600) based on the pre-processed image data, and detect the slope of the communication signal cable (630) to achieve the detachment detection of the communication signal cable (630).

8. The detection system according to claim 7, characterized in that: It is provided with an integrated device (300) externally mounted on a carrier vehicle (100) and a mounting bracket (200) for connecting the integrated device (300) to the carrier vehicle (100), wherein the image acquisition module (342) is mounted inside the integrated device (300), wherein the mounting bracket (200) includes two non-coplanar mounting surfaces, a first mounting surface (211) for connecting to the carrier vehicle (100), and a second mounting surface (212) for connecting to the integrated device (300).

9. The detection system according to claim 7 or 8, characterized in that: The integrated device (300) is provided with a shell portion (310) including an internal cavity, and a mounting groove (321) for connecting with a second mounting surface (212) of the mounting bracket (200) is provided on a bottom plate (320) of the shell portion (310), so that the integrated device (300) can be detachably fixed on the mounting bracket (200).

10. The detection system according to any one of claims 7 to 9, characterized in that: The housing (310) of the integrated device (300) comprises two side panels (330) arranged opposite to each other and a working plate (340) arranged between the two side panels (330); the working plate (340) is provided with a plurality of through holes (341) at intervals, so that the linear array camera (343) and the laser fill light (345) of the image acquisition module (342) can be installed on the through holes (341) in a set manner; wherein the opening position of the through holes (341) on the working plate (340) can be determined based on the field of view angle of the linear array camera (343), so that the imaging of the surface of the tunnel lining (500) except the track area can be achieved while ensuring that the field of view ranges (344) of adjacent linear array cameras (343) are at least partially overlapped.

Images