A ballast track bed disease detection method, device, equipment and storage medium
By combining ground-penetrating radar and deep neural network models, automated detection of defects in ballasted track bed is achieved, solving the problems of low efficiency and poor accuracy in existing technologies and improving the accuracy and adaptability of defect identification.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHENZHEN UNIV
- Filing Date
- 2026-03-11
- Publication Date
- 2026-06-12
Smart Images
Figure CN122199449A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of track bed risk monitoring technology, and in particular to a method, apparatus, equipment and storage medium for detecting defects in ballast track beds. Background Technology
[0002] With the development of railway transportation towards high speed and heavy load, track bed defects (such as compaction and voids) in ballasted tracks have become a key hidden danger threatening operational safety. Achieving accurate and efficient detection of these defects is crucial to ensuring railway safety.
[0003] Currently, track bed condition monitoring mainly relies on manual inspections and simple tool measurements. This method is not only inefficient and consumes a lot of manpower, but the inspection results are also significantly affected by the inspector's experience, weather conditions, lighting, and other external environmental factors. It is highly subjective, makes it difficult to detect hidden defects, and cannot achieve quantitative and systematic evaluation, which is no longer able to meet the high-density and high-safety operation and maintenance requirements of modern railways.
[0004] In recent years, ground-penetrating radar (GPR) technology has been used to acquire images of the interior of railway tracks, but defect identification still relies on visual interpretation by professionals, resulting in low automation. In complex field environments, radar images have low signal-to-noise ratios, making defect feature extraction difficult. Existing methods lack effective intelligent processing tools, and both detection accuracy and efficiency need to be improved. Summary of the Invention
[0005] This invention provides a method, apparatus, equipment, and storage medium for detecting defects in ballasted track beds, enabling automated detection of defects in ballasted track beds. It can accurately identify and locate defects under complex, dirty, and defective conditions, improving detection efficiency and accuracy, and adapting to actual detection environments.
[0006] According to one aspect of the present invention, a method for detecting defects in ballasted track bed is provided. The method includes:
[0007] Acquire the original track bed image of the ballasted track bed to be detected by ground penetrating radar, wherein the original track bed image is a B-scan image;
[0008] The original track bed image is subjected to image standard processing to obtain a standard track bed image, wherein the image standard processing includes at least image denoising processing;
[0009] Based on the standard track bed image and the pre-trained target defect detection model, the track bed defect information of the ballasted track bed is determined, wherein the target defect detection model is a deep neural network model trained based on sample data.
[0010] According to another aspect of the present invention, a device for detecting defects in ballast track bed is provided. The device includes:
[0011] The track bed image acquisition module is used to acquire the original track bed image of the ballasted track bed to be detected by ground penetrating radar, wherein the original track bed image is a B-scan image;
[0012] A standard image processing module is used to perform image standard processing on the original track bed image to obtain a standard track bed image, wherein the image standard processing includes at least image denoising processing;
[0013] The defect information determination module is used to determine the defect information of the ballasted track bed based on the standard track bed image and the pre-trained target defect detection model, wherein the target defect detection model is a deep neural network model trained based on sample data.
[0014] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:
[0015] At least one processor; and
[0016] A memory communicatively connected to the at least one processor; wherein,
[0017] The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the ballast track bed defect detection method according to any embodiment of the present invention.
[0018] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the ballast track bed defect detection method according to any embodiment of the present invention.
[0019] The technical solution of this invention involves acquiring original images of the ballasted track bed to be inspected using ground-penetrating radar. These original images are then subjected to image standardization processing to obtain standard track bed images. Based on these standard track bed images and a pre-trained target defect detection model, automated classification and localization of defects are achieved to determine the track bed defect information. This method accurately reflects the health status of the track bed under different contamination conditions, improves the accuracy of defect identification, and adapts to actual inspection environments.
[0020] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description
[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0022] Figure 1 This is a flowchart of a method for detecting defects in ballast track bed according to an embodiment of the present invention;
[0023] Figure 2 This is a schematic diagram of a gray-level co-occurrence matrix provided according to an embodiment of the present invention;
[0024] Figure 3 This is a schematic diagram of the target disease detection model structure provided by the technical solution of the present invention;
[0025] Figure 4 A flowchart of a method for detecting defects in ballast track bed provided in an embodiment of the present invention;
[0026] Figure 5 This is a schematic diagram of the structure of the sample image generation network provided in an embodiment of the present invention;
[0027] Figure 6 This is a structural diagram of a ballast track bed defect detection device provided according to an embodiment of the present invention;
[0028] Figure 7 This is a schematic diagram of the structure of an electronic device for implementing the method for detecting defects in ballast track bed according to an embodiment of the present invention. Detailed Implementation
[0029] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.
[0030] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0031] Figure 1 This is a flowchart illustrating a method for detecting defects in ballasted track bed according to an embodiment of the present invention. This embodiment is applicable to the rapid detection of defects in ballasted track. The method can be executed by a ballasted track bed defect detection device, which can be implemented in hardware and / or software and can be configured in an electronic device. Figure 1 As shown, the method includes:
[0032] S101. Obtain the original track bed image of the ballasted track bed to be inspected by ground penetrating radar.
[0033] In this invention, the ballasted track bed can refer to a track bed that requires defect detection. The original track bed image can refer to a ground-penetrating radar image of the ballasted track bed.
[0034] In this invention, ground-penetrating radar is used to acquire images of the ballast track bed, thereby obtaining the original track bed image, which is a B-scan image.
[0035] The ground-penetrating radar (GPR) antenna moves at a constant speed along a survey line on the ground. At each measurement point during its movement, the GPR antenna emits an extremely short, high-frequency electromagnetic pulse into the ground. At each measurement point, the antenna receives and records seismic signals reflected from interfaces between different underground media (such as the interface between ballast and roadbed, cavity boundaries, aquifers, etc.).
[0036] As the radar antenna moves continuously along the survey line and records individual seismic trace waveform data from hundreds or thousands of continuous measurement points, the system arranges these one-dimensional seismic trace waveform data to form a two-dimensional B-scan image, which is the original track bed image of the ballast track bed to be inspected.
[0037] S102. Perform image standard processing on the original track bed image to obtain a standard track bed image.
[0038] In this invention, the original track bed images typically contain a large amount of noise and various interferences. Directly extracting features from them would lead to unreliable results or even complete failure. The goal of image standard processing is to transform the original, complex radar images into images that more clearly, realistically, and consistently reflect the underground structure, thereby laying a solid foundation for subsequent feature extraction and defect identification.
[0039] Image standardization processing includes at least image denoising, and may also include interference removal and highlighting of defect features. After standardization processing, the standard track bed image exhibits significantly reduced noise, removal of strong interference bands at the top, visualization of the shallow track bed structure, enhancement of deep signals, and optimization of overall image brightness and contrast, facilitating analysis.
[0040] S103. Based on the standard track bed image and the pre-trained target defect detection model, determine the track bed defect information of the ballasted track bed.
[0041] The target disease detection model is a deep neural network model obtained by training based on sample data.
[0042] Specifically, by using a target defect detection model to analyze the image texture features of standard track bed images, it is possible to determine whether there are track bed defects in ballasted track beds. If so, it is also possible to determine the actual type of track bed defect and its specific location in the image.
[0043] It should be noted that the target defect detection model utilizes the gray-level co-occurrence matrix to perform image texture feature analysis on standard track bed images. Gray-level co-occurrence matrix It is The matrix, where This represents the number of gray levels in the image, and each element in the matrix... Indicates grayscale value and The frequency of pixel pairs occurring under a specific spatial relationship. When constructing the gray-level co-occurrence matrix, it is first necessary to determine the relative positional relationship between two pixels in the image, usually using a displacement vector. To indicate, among which and They are two pixels in and Offset in direction.
[0044] Figure 2 This is a schematic diagram of a gray-level co-occurrence matrix provided according to an embodiment of the present invention. For example... Figure 2As shown, this invention uses four different distances (2, 4, 8, and 16 pixels) to capture texture information of images at different scales; four directions (0°, 45°, 90°, and 135°) are selected to calculate the gray-level co-occurrence matrix to comprehensively analyze the texture features in the image. Then, based on the gray-level co-occurrence matrix, characteristic parameters such as contrast, entropy, energy, inverse difference, and correlation are calculated to analyze the variation patterns of texture features between clean and dirty track beds under different dirt levels. Clean track beds exhibit high contrast, high entropy, and low energy characteristics, while dirty track beds, as the dirt level increases, tend to have a more uniform and smooth texture, exhibiting high energy, high inverse difference, and high correlation characteristics. The target disease detection model can determine the significant features of track beds with different degrees of dirt in ground-penetrating radar imaging based on the above differences, thereby determining the actual type and specific location of track bed defects.
[0045] For example, determining the track bed defect information of the ballasted track bed based on the standard track bed image and the pre-trained target defect detection model includes: inputting the standard track bed image into the target defect detection model for track bed defect detection processing; and determining the location and type of track bed defect in the standard track bed image based on the output of the target defect detection model.
[0046] Specifically, the process of determining track bed defect information using the target defect detection model involves inputting a standard track bed image into the target defect detection model for defect information analysis. After the defect information analysis is confirmed, the target defect detection model outputs the defect information, thus obtaining the track bed defect information for ballasted track.
[0047] Figure 3 This is a schematic diagram of the target disease detection model structure provided by the technical solution of the present invention. For example... Figure 3 As shown, the input standard track bed image undergoes preliminary standardization processing through ConvNormLayer1, extracting basic feature images using convolution, batch normalization, and ReLU activation functions. The basic feature images are then fed into ConvNormLayer2 for further convolution processing to help extract more detailed feature images. Detailed feature images undergo deep feature extraction through multiple UIB_Block modules. Each UIB Block contains GSConv (depth-separable convolution), responsible for extracting multi-level features from the detailed feature images, enhancing the ability to identify local disease features, and obtaining enhanced feature images. The CG-AIFI module analyzes the global context information of the enhanced feature images to help the network capture global patterns in disease areas, improving attention to important areas. After multi-scale feature extraction, the image's features at different scales (such as...) are further processed. ) via GSConv Convolutional layers are fused to enable the network to better identify lesions of different sizes. In the DDG-CCFF module, feature maps are further enhanced through dynamic sampling (DySample) and a deep convolutional module (DDBC3), paying particular attention to detailed regions in the image and optimizing the representation of lesion areas. Finally, the Query Selection module selects the most critical feature regions in the image and passes them to the Decoder & Head module for decoding and classification, ultimately outputting the precise location and type of the lesion.
[0048] The technical solution of this invention involves acquiring original images of the ballasted track bed to be inspected using ground-penetrating radar. These original images are then subjected to image standardization processing to obtain standard track bed images. Based on these standard track bed images and a pre-trained target defect detection model, automated classification and localization of defects are achieved to determine the track bed defect information. This method accurately reflects the health status of the track bed under different contamination conditions, improves the accuracy of defect identification, and adapts to actual inspection environments.
[0049] Based on the above embodiments, the step of performing image standard processing on the original track bed image to obtain a standard track bed image includes: performing direct wave elimination processing on the original track bed image to obtain a transition track bed image; and performing singular value removal processing on the transition track bed image to obtain a standard track bed image.
[0050] In other words, in this invention, image standardization processing of the original track bed image includes direct wave elimination and singular value removal. After direct wave elimination processing of the original track bed image, the resulting image is determined as a transition track bed image. After singular value removal processing of the transition track bed image, a standard track bed image can be obtained. In another embodiment, image standardization processing can be performed by first performing singular value removal processing on the original track bed image, and then performing direct wave elimination processing to obtain a standard track bed image. This invention does not specifically limit the execution order of direct wave elimination processing and singular value removal processing.
[0051] For example, the step of performing direct wave cancellation processing on the original track bed image to obtain a transition track bed image includes: determining the sampling mean of each row of images based on the sampling values of each row of images in the original track bed image; constructing a sampling mean image corresponding to the original track bed image based on the sampling mean; and determining the difference image between the original track bed image and the sampling mean image as the transition track bed image.
[0052] It should be noted that when the electromagnetic waves emitted by ground-penetrating radar propagate downwards, the first signal received by the receiver is the direct wave signal. The direct wave is the strongest signal obtained during the ground-penetrating radar detection process. It comes directly from the signal emitted by the radar and the ground reflection wave, reaching the receiver without being reflected by underground objects. The energy and amplitude of the direct wave are usually much greater than those of the waves reflected by underground objects, and its energy is greater and more uniform.
[0053] This invention employs an averaging elimination method for direct wave elimination. The averaging elimination method leverages the high and relatively uniform energy of direct waves by calculating the average sampling value of each row of the original track bed image. Based on these average sampling values, a sampled mean image is constructed, and this image is subtracted from the original track bed image, thereby achieving effective elimination of direct waves.
[0054] For example, the step of performing singular value removal processing on the transition track bed image to obtain a standard track bed image includes: performing singular value decomposition processing on the transition track bed image to obtain an original left singular matrix, an original singular value matrix, and an original right singular matrix; performing threshold removal processing on the original singular value matrix based on a preset singular threshold to obtain a target singular value matrix; and performing reconstruction processing based on the target singular value matrix, the original left singular matrix, and the original right singular matrix to obtain the standard track bed image.
[0055] In this invention, the singular value removal process mainly involves decomposing the transition track bed image into three matrices, obtaining the original left singular matrix, the original singular value matrix, and the original right singular matrix. Based on a pre-set singular threshold, values in the original singular value matrix exceeding the threshold are removed, and the remaining values are determined as the target singular value matrix. Based on the target singular value matrix, image reconstruction is performed using the original left and right singular matrices to remove larger singular values, effectively suppress the influence of direct waves, remove direct wave components from the radar image, retain more representative underground object reflection signals, and obtain a standard track bed image.
[0056] Figure 4 This is a flowchart illustrating a method for detecting defects in ballasted track bed according to an embodiment of the present invention. This embodiment refines the training process of the target defect detection model based on the aforementioned embodiments. For example... Figure 4 As shown, the method includes:
[0057] S201. Obtain the target track bed sample image and the corresponding label disease results.
[0058] The target track bed sample images can include positive sample images and negative sample images. Positive sample images can be pre-collected images of track bed defects, while negative sample images can be images of track bed without defects. The ratio of positive to negative sample images can be 5:1. Defect labeling results can be obtained by manually annotating the target track bed sample images.
[0059] In this invention, the original track bed image can be used as the target track bed sample image, or the original track bed image can be expanded to obtain a target track bed sample image of a higher order of magnitude.
[0060] For example, the process of obtaining the target track bed sample image includes: obtaining a pre-constructed initial track bed sample image; performing image standard processing on the initial track bed sample image to obtain a standard track bed sample image; and performing sample image augmentation processing based on a pre-constructed sample image generation network and the initial track bed sample image to obtain the target track bed sample image.
[0061] The sample image generation network includes a sample generator and a sample discriminator.
[0062] Specifically, after performing image standardization processing on the initial track bed sample image, a standard track bed sample image is obtained. Based on the sample image generation network, the initial track bed sample image is then subjected to sample image augmentation processing to obtain the target track bed sample image.
[0063] Figure 5 This is a schematic diagram of the structure of a sample image generation network provided according to an embodiment of the present invention. Figure 5 As shown, starting with an input random noise vector, the generator maps the noise into a preliminary radar image through a multi-layer network. During the generation process, a Triplet Attention (TA) mechanism is incorporated, where the generator adaptively weights the feature channels, key regions, and details of the initial track bed sample image through channel, spatial, and detail attention, respectively, thereby improving the quality and quantity of generated images and obtaining the target track bed sample image.
[0064] Furthermore, the generated target track bed sample image is then fed into a discriminator, which evaluates the difference between the generated image and the real image based on Wasserstein distance and gradient penalty, and optimizes the generator through feedback signals. Through adversarial training, the generator continuously improves the detail representation of the image, ultimately generating sample images that are highly similar to the real track bed sample images.
[0065] S202. Input the target track bed sample image into the preset disease detection model for disease detection, and obtain the output disease result based on the output of the preset disease detection model.
[0066] S203. Based on the output disease results and the labeled disease results, determine the training error, and backpropagate the training error to the preset disease detection model to adjust the network parameters in the preset disease detection model.
[0067] Specifically, the training error can be determined based on the training function, the output disease results and labeled disease results of the preset disease detection model, and the training error can be backpropagated to the preset disease detection model. The network parameters in the preset disease detection model can be adjusted, and the above process can be repeated until the model training is completed.
[0068] S204. When the preset convergence condition is met, the training of the preset disease detection model is completed, and the target disease detection model is obtained.
[0069] In this invention, when a preset convergence condition is met, such as when the number of iterations reaches a preset number or the training error converges, it can be determined that the training of the preset disease detection model has ended. At this time, the preset disease detection model that has ended training can be used as the target disease detection model.
[0070] S205. Obtain the original track bed image of the ballasted track bed to be inspected by ground penetrating radar.
[0071] S206. Perform image standard processing on the original track bed image to obtain a standard track bed image.
[0072] S207. Based on the standard track bed image and the pre-trained target defect detection model, determine the track bed defect information of the ballasted track bed.
[0073] The technical solution of this invention involves inputting the target roadbed sample image into a preset disease detection model for disease detection, and obtaining output disease results based on the output of the preset disease detection model. Training error is determined based on the output disease results and the labeled disease results, and this training error is backpropagated to the preset disease detection model to adjust the network parameters. When a preset convergence condition is met, the training of the preset disease detection model is considered complete, and the target disease detection model is obtained. By using the target roadbed sample image and the corresponding labeled disease results for model training, the accuracy of the target disease detection model in detecting roadbed diseases can be guaranteed.
[0074] Figure 6 This is a schematic diagram of a ballast track bed defect detection device provided in an embodiment of the present invention. Figure 6 As shown, the device includes:
[0075] The track bed image acquisition module 301 is used to acquire the original track bed image of the ballasted track bed to be detected by ground penetrating radar, wherein the original track bed image is a B-scan image;
[0076] The standard image processing module 302 is used to perform image standard processing on the original track bed image to obtain a standard track bed image, wherein the image standard processing includes at least image denoising processing;
[0077] The defect information determination module 303 is used to determine the defect information of the ballasted track bed based on the standard track bed image and the pre-trained target defect detection model, wherein the target defect detection model is a deep neural network model trained based on sample data.
[0078] The technical solution of this invention involves acquiring original images of the ballasted track bed to be inspected using ground-penetrating radar. These original images are then subjected to image standardization processing to obtain standard track bed images. Based on these standard track bed images and a pre-trained target defect detection model, automated classification and localization of defects are achieved to determine the track bed defect information. This method accurately reflects the health status of the track bed under different contamination conditions, improves the accuracy of defect identification, and adapts to actual inspection environments.
[0079] Optionally, the disease information determination module 303 is specifically used for:
[0080] The standard track bed image is input into the target defect detection model for track bed defect detection processing;
[0081] Based on the output of the target disease detection model, the location and type of roadbed disease in the standard roadbed image are determined.
[0082] Optionally, the standard image processing module 302 includes:
[0083] The first processing unit is used to perform direct wave elimination processing on the original track bed image to obtain a transition track bed image;
[0084] The second processing unit is used to perform singular value removal processing on the transition track bed image to obtain a standard track bed image.
[0085] Optionally, the first processing unit is specifically used for:
[0086] The average value of each row of images is determined based on the sampled values of each row of images in the original track bed image.
[0087] Based on the sampling mean, construct the sampling mean image corresponding to the original track bed image;
[0088] The difference image between the original track bed image and the sampled mean image is determined as the transition track bed image.
[0089] Optionally, the second processing unit is specifically used for:
[0090] The transition track bed image is subjected to singular value decomposition to obtain the original left singular matrix, the original singular value matrix, and the original right singular matrix.
[0091] Based on a preset singularity threshold, the original singular value matrix is subjected to threshold removal processing to obtain the target singular value matrix.
[0092] The standard track bed image is obtained by reconstructing the target singular value matrix, the original left singular matrix, and the original right singular matrix.
[0093] Optionally, the device further includes a detection model training module. The detection model training module includes:
[0094] The sample image acquisition unit is used to acquire a target roadbed sample image and the corresponding label disease results of the target roadbed sample image;
[0095] The disease result output unit is used to input the target roadbed sample image into a preset disease detection model for disease detection, and obtain the output disease result based on the output of the preset disease detection model;
[0096] The network parameter adjustment unit is used to determine the training error based on the output disease results and the labeled disease results, and to backpropagate the training error to the preset disease detection model to adjust the network parameters in the preset disease detection model.
[0097] The detection model determination unit is used to determine the end of the training of the preset disease detection model and obtain the target disease detection model when the preset convergence condition is met.
[0098] Optionally, the sample image acquisition unit is used for:
[0099] Obtain a pre-constructed initial track bed sample image;
[0100] The initial track bed sample image is subjected to image standardization processing to obtain a standard track bed sample image;
[0101] The target track bed sample image is obtained by performing sample image augmentation processing based on a pre-constructed sample image generation network and the initial track bed sample image. The sample image generation network includes a sample generator and a sample discriminator.
[0102] The ballast track bed defect detection device provided in this embodiment of the invention can execute the ballast track bed defect detection method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.
[0103] Figure 7 A schematic diagram of an electronic device 10, which can be used to implement embodiments of the present invention, is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.
[0104] like Figure 7 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.
[0105] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.
[0106] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, central processing unit (CPU), graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as the method for detecting defects in ballast track bed.
[0107] In some embodiments, the ballast track bed defect detection method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the ballast track bed defect detection method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the ballast track bed defect detection method by any other suitable means (e.g., by means of firmware).
[0108] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.
[0109] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.
[0110] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.
[0111] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).
[0112] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.
[0113] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.
[0114] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.
[0115] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.
Claims
1. A method for detecting defects in ballast track bed, characterized in that, include: Acquire the original track bed image of the ballasted track bed to be detected by ground penetrating radar, wherein the original track bed image is a B-scan image; The original track bed image is subjected to image standard processing to obtain a standard track bed image, wherein the image standard processing includes at least image denoising processing; Based on the standard track bed image and the pre-trained target defect detection model, the track bed defect information of the ballasted track bed is determined, wherein the target defect detection model is a deep neural network model trained based on sample data.
2. The method according to claim 1, characterized in that, The step of determining the track bed defect information of the ballasted track bed based on the standard track bed image and the pre-trained target defect detection model includes: The standard track bed image is input into the target defect detection model for track bed defect detection processing; Based on the output of the target disease detection model, the location and type of roadbed disease in the standard roadbed image are determined.
3. The method according to claim 1, characterized in that, The step of performing image standardization processing on the original track bed image to obtain a standard track bed image includes: The original track bed image is subjected to direct wave elimination processing to obtain a transition track bed image; The transition track bed image is subjected to singular value removal processing to obtain the standard track bed image.
4. The method according to claim 3, characterized in that, The step of performing direct wave elimination processing on the original track bed image to obtain a transition track bed image includes: The average value of each row of images is determined based on the sampled values of each row in the original track bed image. Based on the sampling mean, construct the sampling mean image corresponding to the original track bed image; The difference image between the original track bed image and the sampled mean image is determined as the transition track bed image.
5. The method according to claim 3, characterized in that, The step of performing singular value removal processing on the transition track bed image to obtain a standard track bed image includes: The transition track bed image is subjected to singular value decomposition to obtain the original left singular matrix, the original singular value matrix, and the original right singular matrix. Based on a preset singularity threshold, the original singular value matrix is subjected to threshold removal processing to obtain the target singular value matrix. The standard track bed image is obtained by reconstructing the target singular value matrix, the original left singular matrix, and the original right singular matrix.
6. The method according to claim 1, characterized in that, The training process of the target disease detection model includes: Obtain the target track bed sample image and the corresponding labeled disease results; The target track bed sample image is input into a preset disease detection model for disease detection, and the output disease result is obtained based on the output of the preset disease detection model; The training error is determined based on the output disease results and the labeled disease results, and the training error is backpropagated to the preset disease detection model to adjust the network parameters in the preset disease detection model. When the preset convergence condition is met, the training of the preset disease detection model is considered complete, and the target disease detection model is obtained.
7. The method according to claim 6, characterized in that, The process of acquiring the target track bed sample image includes: Obtain a pre-constructed initial track bed sample image; The initial track bed sample image is subjected to image standardization processing to obtain a standard track bed sample image; The target track bed sample image is obtained by performing sample image augmentation processing based on a pre-constructed sample image generation network and the initial track bed sample image. The sample image generation network includes a sample generator and a sample discriminator.
8. A device for detecting defects in ballast track bed, characterized in that, include: The track bed image acquisition module is used to acquire the original track bed image of the ballasted track bed to be detected by ground penetrating radar, wherein the original track bed image is a B-scan image; A standard image processing module is used to perform image standard processing on the original track bed image to obtain a standard track bed image, wherein the image standard processing includes at least image denoising processing; The defect information determination module is used to determine the defect information of the ballasted track bed based on the standard track bed image and the pre-trained target defect detection model, wherein the target defect detection model is a deep neural network model trained based on sample data.
9. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the method for detecting defects in ballast track bed as described in any one of claims 1-7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that are used to cause a processor to execute the method for detecting defects in ballast track bed as described in any one of claims 1-7.