Bridge cable disease detection method and system based on image recognition

Abstract

The invention relates to the technical field of bridge operation and maintenance detection, in particular to a bridge cable disease detection method and system based on image recognition, and the method comprises the steps: collecting the multi-source image data and environmental parameter monitoring data of a bridge cable; performing multi-scale feature fusion processing on the multi-source image data to obtain a cable fusion feature tensor; considering the environmental parameter monitoring data, performing environmental elimination on each feature in the inhaul cable fusion feature tensor, and obtaining an inhaul cable fusion elimination feature tensor; inputting the inhaul cable fusion eliminating feature tensor into a preset composite disease evaluation model to obtain an inhaul cable structure comprehensive disease index; the comprehensive disease index of the inhaul cable structure is compared with a preset safety threshold value, and a graded early warning signal is generated; the system can timely and accurately reflect the overall health condition of the inhaul cable, gives an alarm when the inhaul cable damage is still in a controllable stage, and effectively prevents major safety accidents.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge operation and maintenance detection, and in particular to a bridge cable disease detection method and system based on image recognition. Background Art

[0002] As an important part of modern transportation infrastructure, the safety and durability of bridges are directly related to the smoothness of transportation and the safety of life and property. Bridge cables are key load-bearing components in structures such as cable-stayed bridges and suspension bridges, and their integrity is crucial to the overall performance and stability of the bridge. However, during their long-term service, cables are subject to the interaction of complex environmental conditions and dynamic loads, and often suffer from diseases, which will significantly reduce their bearing capacity and service life, and may even cause major safety accidents.

[0003] Existing methods for detecting bridge cable defects mainly rely on manual inspections and simple instrument inspections, such as visual inspection, paint inspection, ultrasonic inspection, etc. Although the existing methods have certain effectiveness, they have limitations such as low efficiency, poor accuracy, and high cost when facing complex environments and large-scale bridge cable networks. For example, visual inspection relies on the experience and judgment of the inspectors, which is easily affected by subjective factors and difficult to detect hidden defects. Although ultrasonic testing can detect some defects inside the cable, its detection efficiency is low and it is difficult to meet the detection needs of long-distance and large-scale cables. In addition, most of the existing detection methods can only evaluate a certain defect, lack a comprehensive quantitative evaluation method for the comprehensive defects of the cable structure, and cannot give a timely and accurate judgment on the overall health status of the cable and the corresponding early warning. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a bridge cable defect detection method and system based on image recognition, which can timely and accurately reflect the overall health status of the cable, issue an alarm when the cable disease is still in a controllable stage, and effectively prevent major safety accidents.

[0005] In a first aspect, the present invention provides a bridge cable disease detection method based on image recognition, the method comprising: Collect multi-source image data and environmental parameter monitoring data of bridge cables; Performing multi-scale feature fusion processing on the multi-source image data to obtain a Lasso fusion feature tensor; Considering the environmental parameter monitoring data, performing environmental elimination on each feature in the Lasso fusion feature tensor to obtain a Lasso fusion elimination feature tensor; Inputting the cable fusion elimination characteristic tensor into a preset composite disease assessment model to obtain a comprehensive disease index of the cable structure; The comprehensive disease index of the cable structure is compared with a preset safety threshold to generate a graded early warning signal.

[0006] Furthermore, the multi-source image data includes visible light band image data, infrared thermal imaging data and three-dimensional laser point cloud data.

[0007] Furthermore, the cable fusion feature tensor includes surface crack distribution features, cable internal corrosion features and cross-sectional geometric distortion features; wherein the surface crack distribution features are extracted based on visible light band image data through a spatial domain neural network model; the cable internal corrosion features are analyzed based on infrared thermal imaging data through a frequency domain transformation algorithm; and the cross-sectional geometric distortion features are obtained based on three-dimensional laser point cloud data through three-dimensional point cloud matching calculation.

[0008] Furthermore, the process of acquiring the surface crack distribution characteristics includes: Preprocessing the visible light band image data, wherein the preprocessing method includes image denoising, image enhancement and image segmentation; Collect visible light image data containing cracks on the cable surface and build a spatial domain neural network model; The preprocessed visible light band image is input into the spatial domain neural network model to obtain the surface crack distribution characteristics.

[0009] Furthermore, the process of obtaining the internal corrosion characteristics of the cable includes: Preprocessing the infrared thermal imaging data, wherein the preprocessing steps include image enhancement, denoising, and temperature calibration and normalization; The pre-processed infrared thermal imaging data is converted from the time domain to the frequency domain using the fast Fourier transform algorithm; Perform filtering on the frequency domain data after fast Fourier transformation; The internal corrosion characteristics of the cable are extracted from the filtered frequency domain data.

[0010] Furthermore, the process of acquiring the cross-sectional geometric distortion feature includes: Use a 3D laser scanner to scan the bridge cables and obtain 3D laser point cloud data of the cables; Preprocessing the collected original 3D laser point cloud data; the preprocessing steps include removing noise points, normalizing data, and registering point clouds; According to the spatial position and shape of the cable, the cross-sectional point cloud of the cable is extracted from the pre-processed three-dimensional point cloud data; Matching and calculating the extracted cable section point cloud with the preset cable standard section model; According to the matching results, the geometric distortion characteristics of the cable section are calculated.

[0011] Furthermore, the graded warning signals include normal, mild warning, moderate warning, severe warning and critical warning.

[0012] On the other hand, the present application also provides a bridge cable disease detection system based on image recognition, the system comprising: Data acquisition module, used to collect multi-source image data of bridge cables and environmental parameter monitoring data; A feature fusion processing module, used for performing multi-scale feature fusion processing on the multi-source image data to obtain a Lasso fusion feature tensor; An environment elimination processing module, considering the environment parameter monitoring data, performs environment elimination on each feature in the Lasso fusion feature tensor to obtain a Lasso fusion elimination feature tensor; A composite disease assessment module, used for inputting the cable fusion elimination feature tensor into a preset composite disease assessment model to obtain a comprehensive disease index of the cable structure; The early warning generation module is used to compare the comprehensive disease index of the cable structure with a preset safety threshold to generate a graded early warning signal.

[0013] In a third aspect, the present application provides an electronic device, comprising a bus, a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the transceiver, the memory, and the processor are connected via the bus, and the computer program, when executed by the processor, implements the steps of any one of the above methods.

[0014] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps in any one of the above-mentioned methods when executed by a processor.

[0015] Compared with the prior art, the invention has the following beneficial effects: the method collects multi-source image data such as visible light band images, infrared thermal imaging and three-dimensional laser point cloud, each of which provides detailed information on the surface, interior and geometric morphology of the cable; through multi-scale feature fusion processing, this information is integrated into a cable fusion feature tensor, which comprehensively reflects the various disease characteristics of the cable; the surface crack distribution characteristics are extracted from the visible light band images using a spatial domain neural network model, the internal corrosion characteristics of the cable are analyzed from the infrared thermal imaging data using a frequency domain transformation algorithm, and the cross-sectional geometric distortion characteristics are obtained from the three-dimensional laser point cloud data using three-dimensional point cloud matching calculation; the accuracy of feature extraction is improved, making disease detection more accurate and reliable; considering environmental parameter monitoring data, Perform environmental elimination processing on each feature in the cable fusion feature tensor; effectively reduce the impact of environmental factors such as temperature, humidity, wind force on the disease detection results, and improve the stability and reliability of detection; input the cable fusion elimination feature tensor into the preset composite disease assessment model to obtain the cable structure comprehensive disease index; this index comprehensively considers the various disease characteristics of the cable, realizes a comprehensive quantitative assessment of the comprehensive diseases of the cable structure, and provides more comprehensive information support for operation and maintenance personnel; compares the cable structure comprehensive disease index with the preset safety threshold to generate a graded warning signal; the warning signal is graded according to the severity of the disease, so that the operation and maintenance personnel can quickly judge the health of the cable and take corresponding maintenance measures, thereby improving the response efficiency and safety; Multi-source image data acquisition covers visible light band images, infrared thermal imaging data and three-dimensional laser point cloud data, capturing cable status information from multiple dimensions; environmental parameter monitoring data supplements external influencing factors, providing a comprehensive and accurate basis for subsequent analysis; multi-scale feature fusion processing integrates the features of different types of data to generate a cable fusion feature tensor; features such as surface crack distribution, internal corrosion and cross-sectional geometric distortion are no longer isolated, but are fused to form a feature set that comprehensively reflects the health status of the cable; the understanding of cable diseases is no longer limited to a single dimension, but starts from the perspective of overall structural integrity and disease correlation, comprehensively grasping the actual status of the cable, greatly improving the cognitive ability of complex diseases; based on The cable fusion feature tensor is environmentally eliminated based on the environmental parameter monitoring data to remove the interference of environmental factors on the cable characteristics and obtain the cable fusion elimination feature tensor; this process makes the disease characteristics more pure and avoids misjudgment caused by environmental factors; the cable fusion elimination feature tensor is input into the composite disease assessment model to obtain the comprehensive disease index of the cable structure, and compared with the preset safety threshold to generate a graded early warning signal; from multi-source data collection to final early warning, the entire process forms a closed loop, realizing comprehensive automated processing of the cable health status from detection to evaluation to early warning; it can timely and accurately reflect the overall health status of the cable, and issue an alarm when the cable disease is still in a controllable stage, effectively preventing major safety accidents and ensuring the safe operation of the bridge. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a flow chart of the present invention; Figure 2 It is a structural diagram of the bridge cable defect detection system based on image recognition. DETAILED DESCRIPTION

[0017] In the description of this application, those skilled in the art should know that this application can be implemented as a method, an apparatus, an electronic device, and a computer-readable storage medium. Therefore, this application can be specifically implemented in the following forms: complete hardware, complete software (including firmware, resident software, microcode, etc.), a combination of hardware and software. In addition, in some embodiments, this application can also be implemented in the form of a computer program product in one or more computer-readable storage media, and the computer-readable storage medium contains computer program code.

[0018] The above-mentioned computer-readable storage medium may adopt any combination of one or more computer-readable storage media. Computer-readable storage media include: electrical, magnetic, optical, electromagnetic, infrared or semiconductor systems, devices or devices, or any combination of the above. More specific examples of computer-readable storage media include: portable computer disks, hard disks, random access memories, read-only memories, erasable programmable read-only memories, flash memories, optical fibers, optical disc read-only memories, optical storage devices, magnetic storage devices, or any combination of the above. In the present application, computer-readable storage media can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device, or device.

[0019] The acquisition, storage, use, and processing of data in the technical solution of this application comply with the relevant provisions of national laws.

[0020] The present application describes the provided methods, devices, and electronic devices through flowcharts and / or block diagrams.

[0021] It should be understood that each box in the flowchart and / or block diagram and the combination of boxes in the flowchart and / or block diagram can be implemented by computer-readable program instructions. These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer or other programmable data processing device to produce a machine, and these computer-readable program instructions are executed by a computer or other programmable data processing device to produce a device that implements the functions / operations specified by the boxes in the flowchart and / or block diagram.

[0022] These computer-readable program instructions may also be stored in a computer-readable storage medium that enables a computer or other programmable data processing device to work in a specific manner. In this way, the instructions stored in the computer-readable storage medium produce an instruction device product including functions / operations specified in the blocks in the flowchart and / or block diagram.

[0023] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device, so that a series of operational steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby enabling the instructions executed on the computer or other programmable data processing apparatus to provide a process for implementing the functions / operations specified in the blocks in the flowchart and / or block diagram.

[0024] The present application is described below in conjunction with the drawings in the present application.

[0025] Embodiment 1: Figure 1 As shown, the bridge cable disease detection method based on image recognition of the present invention specifically includes the following steps: S1. Collect multi-source image data and environmental parameter monitoring data of bridge cables; the multi-source image data includes visible light band image data, infrared thermal imaging data and three-dimensional laser point cloud data; Visible light band image data acquisition: Use a high-resolution visible light camera. For large bridge cables, use a professional-grade visible light camera mounted on a drone. The drone can flexibly fly to various parts of the cable to achieve no-dead-angle shooting. At the same time, install monitoring cameras at fixed positions on the bridge to conduct long-term continuous monitoring of the cable in order to record the dynamic changes in the surface condition of the cable. When collecting data, ensure sufficient lighting conditions and choose to shoot in the morning or afternoon on sunny days to avoid backlight affecting the image quality. For the key parts of the cable, focus on shooting to ensure that the image can clearly show whether there are cracks, coating peeling and other signs of disease in these parts. In addition, by setting different shooting angles and distances, multiple sets of images can be obtained to provide more comprehensive data support for subsequent image analysis. Infrared thermal imaging data collection: A high-precision infrared thermal imager with a highly sensitive thermal sensor is used to detect tiny temperature differences on the surface of the cable. The infrared thermal imager is also used on a drone to quickly scan the entire cable. For some difficult-to-reach locations, remote-controlled infrared thermal imaging equipment is used for monitoring. Since the temperature is greatly affected by environmental factors, data collection is carried out during periods when the temperature is relatively stable. During the data collection process, the infrared thermal imager must be accurately calibrated to ensure the accuracy of temperature measurement. At the same time, according to the material and structural characteristics of the cable, the parameters of the thermal imager are reasonably set to obtain infrared thermal images that best reflect the internal corrosion status of the cable. 3D laser point cloud data acquisition: Using a 3D laser scanner, the 3D spatial coordinate information of the cable can be quickly obtained to generate accurate point cloud data; the 3D laser scanner is installed on a movable platform and slowly moved along the direction of the cable for scanning; for bridge cables with complex structures, a portable 3D laser scanner is used to perform additional scanning of local details; before scanning, the scanning area needs to be reasonably planned to determine the scanning range and accuracy requirements; during the scanning process, the relative position between the scanner and the cable must be stable to avoid data errors caused by shaking; at the same time, by setting appropriate scanning parameters, point cloud data of sufficient density can be obtained so that the cross-sectional geometry and size of the cable can be accurately restored later; Environmental parameter monitoring data collection includes meteorological parameter collection and bridge structure response parameter collection; Meteorological parameter collection: Meteorological stations are installed around the bridges, equipped with temperature sensors, humidity sensors, wind speed sensors, rainfall sensors and other equipment to monitor ambient temperature, relative humidity, wind speed, rainfall and other meteorological parameters in real time; data from the meteorological stations are uploaded to the data processing center in real time via wireless transmission; the installation location of the meteorological station should be selected in an open, well-ventilated area that can represent the surrounding environment of the bridge, avoiding the influence of obstructions such as buildings and trees; meteorological sensors should be calibrated and maintained regularly to ensure the accuracy and reliability of the collected data; Acquisition of bridge structure response parameters: Install strain gauges, accelerometers and other equipment on bridge cables and key structural parts to monitor the strain changes of cables under dynamic loads and the vibration response of bridge structures; the collected data is transmitted to the data acquisition instrument via wired or wireless means, and then summarized to the data processing center; the installation position of the sensor should be reasonably arranged according to the mechanical properties of the bridge structure and the parts prone to diseases, to ensure that the key response information of the structure can be accurately captured; during the installation process, it is necessary to ensure that the sensor is tightly connected to the structure to avoid looseness or poor contact; at the same time, the collected data is filtered and processed in real time to remove noise interference and extract effective information.

[0026] In this step, by collecting multi-source image data such as visible light band images, infrared thermal imaging and three-dimensional laser point cloud, the surface condition, internal corrosion and geometric shape of the bridge cables can be reflected in all directions and from multiple angles, ensuring the comprehensiveness and accuracy of disease detection; the use of drones equipped with professional-grade visible light cameras and infrared thermal imagers can achieve fast and blind-angle shooting and scanning of large bridge cables, greatly improving the efficiency of data collection; the use of three-dimensional laser scanners can quickly obtain the three-dimensional spatial coordinate information of the cables and generate accurate point cloud data, providing efficient data processing methods for subsequent disease analysis and evaluation; the installation of meteorological stations and sensors realizes real-time monitoring of the surrounding environment and structural response of the bridge, providing timely and accurate data support for bridge operation and maintenance management; the reliability and stability of three-dimensional laser point cloud data are guaranteed by reasonably planning the scanning area, setting scanning parameters and ensuring the stability of the relative position between the scanner and the cable; regular calibration and maintenance of meteorological sensors, as well as reasonable layout and close connection of sensors, ensure the accuracy and reliability of collected data; real-time filtering and processing of collected data removes noise interference, extracts effective information, and further improves the reliability and stability of data.

[0027] S2, performing multi-scale feature fusion processing on the multi-source image data to obtain a Lasso fusion feature tensor; The cable fusion feature tensor includes surface crack distribution features, cable internal corrosion features and cross-sectional geometric distortion features; wherein the surface crack distribution features are extracted based on visible light band image data through a spatial domain neural network model; the cable internal corrosion features are analyzed based on infrared thermal imaging data through a frequency domain transformation algorithm; the cross-sectional geometric distortion features are obtained based on three-dimensional laser point cloud data through three-dimensional point cloud matching calculation; Pre-process the collected visible light band image data, infrared thermal imaging data and 3D laser point cloud data; perform denoising and contrast enhancement on the visible light band image data to highlight the signs of defects such as cracks and coating peeling on the cable surface; perform temperature calibration and heat map generation on the infrared thermal imaging data to accurately reflect the temperature distribution on the cable surface and the internal corrosion condition; perform denoising and point cloud registration on the 3D laser point cloud data to obtain accurate 3D spatial coordinate information of the cable; The process of acquiring the surface crack distribution characteristics includes: Preprocessing the visible light band image data, the preprocessing method includes image denoising, the collected visible light image will be disturbed by environmental noise, equipment noise, etc., affecting the crack feature extraction; using Gaussian filtering, median filtering and other algorithms to remove noise in the image; the preprocessing method also includes image enhancement, in order to more clearly highlight the surface cracks of the cable, using histogram equalization, contrast-limited adaptive histogram equalization and other methods to enhance the image; histogram equalization redistributes the grayscale values ​​of image pixels to make the image grayscale distribution more uniform, enhance the overall contrast, and make the originally inconspicuous cracks easier to observe; contrast-limited adaptive histogram equalization performs histogram equalization on local areas to avoid over-enhancement or detail loss problems that may occur during global enhancement, and can better highlight the crack characteristics of the local area on the cable surface; The preprocessing method also includes image segmentation, which separates the cable from the complex background environment so as to focus on the analysis of the surface features of the cable; using methods such as threshold segmentation, edge detection algorithm or semantic segmentation model based on deep learning; based on threshold segmentation, a threshold is set according to the grayscale difference between the cable and the background, and the image is divided into a cable area and a background area; the edge detection algorithm calculates the image gradient, finds the edge strength and direction, and can accurately detect the cable edge, providing a basis for subsequent segmentation; the semantic segmentation model based on deep learning can accurately segment the cable area through training with a large amount of labeled data, and can effectively identify the cable boundary even in a complex background; A large amount of visible light image data containing cracks on the cable surface is collected to build a spatial domain neural network model; a convolutional neural network architecture is selected, which includes a module consisting of multiple convolutional layers and pooling layers alternating with each other, as well as a fully connected layer; first, multiple convolutional layer modules are constructed, each module consists of two to three convolutional layers, and the convolutional layer uses a small 3×3 convolution kernel to extract features at different levels of the cable surface through multiple convolution operations; then the pooling layer is connected, and the maximum pooling is used to reduce the resolution of the feature map without losing key features, thereby reducing the amount of calculation; after multiple convolutions- After pooling, the feature map is expanded and connected to the fully connected layer, the extracted features are comprehensively processed, and the surface crack distribution feature vector is output; a large number of visible light images containing cracks on the cable surface are collected and annotated; the annotation content includes information such as the location, length, width, and direction of the cracks; the annotated images are divided into training set, validation set, and test set, with a ratio of 70%, 15%, and 15%; data enhancement operations are performed on the training set images to improve the generalization ability of the model; the constructed convolutional neural network model is trained using the annotated data; a suitable loss function, such as the cross entropy loss function, is selected to measure the difference between the model prediction results and the true annotations, and the model parameters are adjusted; during the training process, appropriate hyperparameters such as the learning rate and number of iterations are set; in each training batch, the model forward propagates to calculate the prediction results, calculates the error through the loss function, and then backpropagates to update the model parameters, and continuously iterates the training until the performance indicators of the model on the validation set reach a stable and satisfactory level; The preprocessed visible light band image is input into the trained spatial domain neural network model to obtain the surface crack distribution characteristics; the image is calculated by the convolution layer, pooling layer and other network layers in turn; in the convolution layer, different convolution kernels slide on the image and perform convolution operations with the local area of ​​the image to extract various local features; the pooling layer downsamples the feature map output by the convolution layer to retain the main features while reducing the amount of data; after calculation by multiple layers of network layers, a feature vector is finally output in the fully connected layer, which contains the distribution characteristics of the cracks on the cable surface, such as the location, shape, density and other information of the cracks. The process of obtaining the internal corrosion characteristics of the cable includes: The infrared thermal imaging data is preprocessed, and the preprocessing steps include image enhancement and denoising. The collected infrared thermal imaging data may be interfered by noise, resulting in inaccurate temperature information. Gaussian filtering, median filtering and other algorithms are used to remove image noise. Gaussian filtering performs weighted averaging of neighborhood pixels based on Gaussian distribution, which can effectively smooth the image, reduce the impact of random noise, and ensure the accuracy of temperature data. At the same time, the image contrast is enhanced by histogram equalization or local histogram equalization, making the temperature difference on the cable surface more obvious and highlighting the abnormal temperature area that may be caused by internal corrosion. Histogram equalization redistributes the image grayscale value to make the temperature distribution more uniform and improve the overall visibility. Local Histogram equalization optimizes local areas to avoid detail loss caused by global enhancement and better highlight the local temperature changes of the cable; the preprocessing step also includes temperature calibration and normalization. Due to environmental factors and the characteristics of the equipment itself, the temperature measured by the infrared thermal imager will have deviations; the collected data needs to be calibrated according to the equipment calibration parameters and the known standard temperature source to ensure accurate temperature measurement; then, the calibrated temperature data is normalized to a specific range to facilitate subsequent data analysis and processing; this makes the data collected at different times and in different environments comparable, eliminates the interference caused by the difference in absolute temperature values, and focuses more on the relationship between relative temperature changes and internal corrosion; The pre-processed infrared thermal imaging data is converted from the time domain to the frequency domain using the fast Fourier transform algorithm. In the time domain, the data uses time or spatial position as a variable, presenting the variation of the cable surface temperature with position. After the fast Fourier transform, the data uses frequency as a variable in the frequency domain, revealing the frequency components of the temperature variation. The temperature anomaly caused by the internal corrosion of the cable often corresponds to specific frequency characteristics, which can be more clearly identified through frequency domain analysis. The frequency domain data after fast Fourier transformation is filtered. A bandpass filter is used to set the appropriate passband frequency according to the frequency range of temperature changes caused by internal corrosion of the cable. High-frequency noise and low-frequency background interference are removed, and the mid-frequency band frequency information related to internal corrosion is retained. High-frequency noise is caused by interference from electronic components of the equipment or instantaneous temperature fluctuations in the environment, while low-frequency background reflects the overall ambient temperature trend, which are unrelated to the internal corrosion characteristics of the cable. These interferences are removed by filtering, and the key frequency characteristics corresponding to internal corrosion are highlighted, providing purer data for subsequent feature extraction. The internal corrosion characteristics of the cable are extracted from the filtered frequency domain data; the characteristic parameters include the energy value and power spectrum density of a specific frequency band; the energy value reflects the total intensity of the signal in the frequency band, and the power spectrum density describes the distribution of the signal power at different frequencies; for the internal corrosion area, its energy value and power spectrum density in a specific frequency band usually deviate from the normal range; by calculating the energy value and power spectrum density, the internal corrosion characteristics of the cable can be quantified, providing a numerical basis for subsequent evaluation; a mapping relationship between the internal corrosion characteristics of the cable and the frequency domain parameters is established, and the characteristic parameters are classified using a machine learning classification algorithm; first, a large amount of data is collected The infrared thermal imaging data of cables with different degrees of internal corrosion and normal cables are collected, and their frequency domain characteristic parameters are annotated; then, the annotated data are divided into a training set and a test set, and the classification algorithm is trained using the training set to optimize the model parameters so that it can accurately identify different characteristic patterns, that is, distinguish normal cables from cables with internal corrosion, and judge the degree of corrosion; the model performance is verified on the test set to ensure its accuracy and reliability; in this way, according to the extracted frequency domain characteristic parameters, the trained classification model is used to judge whether the cable has internal corrosion and the severity of the corrosion, thereby obtaining the internal corrosion characteristics of the cable; The process of acquiring the cross-sectional geometric distortion feature includes: Use a 3D laser scanner to scan the bridge cables and obtain the 3D laser point cloud data of the cables. During the acquisition process, the scanner parameters need to be set according to factors such as the length and diameter of the cables and the on-site environment to ensure that complete and accurate point cloud data is obtained. At the same time, in order to improve the accuracy and reliability of the data, it is necessary to scan from multiple different angles and then splice and merge the data. Preprocess the collected original 3D laser point cloud data; the preprocessing steps include removing noise points, removing isolated points and abnormal points in the point cloud through filtering algorithms such as Gaussian filtering and median filtering; data normalization, normalizing the coordinate values ​​of the point cloud data to a certain range for subsequent calculation and processing; point cloud registration, for data collected from multiple angles, point cloud registration is required to unify the point cloud data from different perspectives into the same coordinate system; According to the spatial position and shape of the cable, the cross-sectional point cloud of the cable is extracted from the preprocessed 3D point cloud data; by setting appropriate thresholds and geometric constraints, the point cloud belonging to the cross-sectional area of ​​the cable is separated from other background point clouds; The extracted cable section point cloud is matched with the pre-established cable standard section model; the standard section model is an ideal cable section generated according to the design drawing or theoretical model; the purpose of the matching calculation is to find the best matching relationship between the current section point cloud and the standard section model; According to the matching results, the geometric distortion characteristics of the cable section are calculated; In this step, by combining visible light band image data, infrared thermal imaging data and three-dimensional laser point cloud data, this step can comprehensively and deeply extract the surface crack distribution characteristics, internal corrosion characteristics and cross-sectional geometric distortion characteristics of the cable, which together constitute a comprehensive description of the health status of the cable; for different types of disease characteristics, this step adopts different preprocessing methods and feature extraction algorithms, which can more clearly highlight the cracks and other disease signs on the cable surface; for infrared thermal imaging data, temperature calibration, heat map generation and frequency domain analysis are used to accurately reflect the corrosion status inside the cable, significantly improving the accuracy of disease detection; this step fully considers the influence of environmental factors on the detection results in the feature extraction process, so that this method can maintain a high detection accuracy under different environments. Accuracy; this step fuses the disease features of different sources and scales to obtain the cable fusion feature tensor; it not only retains the independence of each feature, but also realizes the complementarity and enhancement of information through feature fusion; it can more comprehensively reflect the health status of the cable and improve the reliability and accuracy of disease detection; through the processing of this step, a cable fusion feature tensor containing surface crack distribution features, internal corrosion features and cross-sectional geometric distortion features is obtained; this feature tensor will be used as input data for the subsequent comprehensive disease assessment model, providing strong support for the quantitative assessment of the health status of the cable and graded early warning; the multi-scale feature fusion processing in this step has important application value in the bridge cable disease detection method, and can significantly improve the accuracy and reliability of disease detection.

[0028] S3, considering the environmental parameter monitoring data, performing environmental elimination on each feature in the Lasso fusion feature tensor to obtain a Lasso fusion elimination feature tensor; The method for performing environmental elimination on each feature in the Lasso fusion feature tensor comprises: Conduct a detailed analysis of the collected environmental parameter monitoring data; clarify the impact mechanism of each environmental parameter on each feature in the cable fusion feature tensor; for example, temperature changes may affect the thermal expansion of the cable material, thereby affecting the cross-sectional geometric distortion characteristics; humidity is related to the surface crack distribution characteristics and internal corrosion characteristics, because a high humidity environment tends to accelerate the corrosion process, and also make surface cracks more obvious or generate new cracks; According to the relationship between environmental parameters and cable characteristics, a corresponding environmental impact model is established; the specific form of the model can be determined by theoretical analysis, experimental research or reference to relevant literature; the model is used to quantify the degree of influence of environmental factors on various characteristics of the cable; According to the environmental impact model, the various features in the cable fusion feature tensor are adjusted; for the surface crack distribution feature, if it is found that high humidity may make the cracks more visually obvious, the feature is corrected according to the humidity crack visual enhancement model to remove the false enhancement caused by humidity factors; for the internal corrosion feature of the cable, according to the influence model of environmental parameters such as temperature and humidity on the corrosion rate, the corrosion feature obtained by the frequency domain transformation algorithm is adjusted to eliminate the corrosion feature deviation caused by environmental factors; for the cross-sectional geometric distortion feature, considering the influence of thermal expansion or contraction caused by temperature and cable vibration caused by wind on the geometric dimension measurement, the feature is corrected using the corresponding environmental geometry change model to obtain a feature that more accurately reflects the real geometric distortion of the cable; The various features adjusted by environmental factors are fused to form a cable fusion elimination feature tensor.

[0029] In this step, by analyzing the environmental parameter monitoring data in detail and clarifying the influence mechanism of each environmental parameter on the cable characteristics, this step can quantify the interference degree of environmental factors on the various characteristics of the cable; then, according to the environmental impact model, the various characteristics in the cable fusion feature tensor are adjusted, which can significantly reduce the impact of environmental factors on the detection results, thereby improving the accuracy of cable disease detection; the environmental elimination process not only considers the impact of a single environmental factor on the cable characteristics, but also integrates the interaction between multiple environmental factors; by establishing and applying the environmental impact model, this step can more comprehensively understand and correct the impact of environmental factors on the cable characteristics, thereby enhancing the reliability and consistency of the detection data; traditional manual inspections and simple instrument detection methods are ineffective in the face of complex environments and large-scale bridge cable networks. When the cable is detected, there are problems such as low efficiency and poor accuracy. This step, by introducing environmental elimination technology, can automatically adjust the cable characteristics, reduce dependence on manual intervention, and thus improve detection efficiency. The cable fusion feature tensor processed by environmental elimination more accurately reflects the true state of the cable. This provides more reliable data input for the subsequent disease assessment model, enabling the model to more accurately assess the health status of the cable and generate graded early warning signals, providing intelligent support for the operation and maintenance management of the bridge. By improving detection accuracy and efficiency, this step helps to reduce unnecessary maintenance and repair work, thereby reducing the overall cost of bridge operation and maintenance. At the same time, the intelligent early warning system can promptly detect potential disease problems, avoid accidents, and further ensure the safety and durability of the bridge.

[0030] S4, inputting the cable fusion elimination characteristic tensor into a preset composite disease assessment model to obtain a comprehensive disease index of the cable structure; After the cable fusion elimination feature tensor is input into the composite disease assessment model, the model performs calculations based on its predefined algorithm and structure; the model performs multiple rounds of feature screening and combination on various features in the input cable fusion elimination feature tensor; the input data is divided and predicted based on different feature subsets, and the final output is determined by majority voting or average prediction value; in this process, the model uses the trained decision rules to conduct a comprehensive analysis from multiple dimensions such as the surface crack distribution characteristics, internal corrosion characteristics, and cross-sectional geometric distortion characteristics of the cable to evaluate the degree of disease of the cable structure; Inside the model, the input features are further fused to extract higher-level feature representations; Based on the fused features, the model outputs a value representing the comprehensive damage degree of the cable structure through calculation and transformation, namely, the comprehensive damage index of the cable structure. In order to ensure the accuracy and reliability of the model, an independent validation data set is used to validate the model; by comparing the consistency between the model prediction results and the actual disease situation, the performance of the model is evaluated, and necessary adjustments and optimizations are made; The calculated comprehensive damage index of the cable structure can be used to evaluate the overall health of the cables and provide decision support for bridge operation and maintenance management.

[0031] In this step, the model can conduct a comprehensive analysis from multiple dimensions such as the surface crack distribution characteristics, internal corrosion characteristics, and cross-sectional geometric distortion characteristics of the cable, comprehensively evaluate the degree of damage to the cable structure, and improve the accuracy and comprehensiveness of damage detection; within the model, the input features are further fused and processed to extract higher-level feature representations; this helps the model to better understand and identify the damage characteristics of the cable, thereby improving the accuracy of the prediction; the model uses the trained decision rules for calculations, and through multiple rounds of feature screening and combination, as well as division and prediction based on different feature subsets, the output is finally determined by majority voting or average prediction value; this improves the stability and reliability of the model; and uses an independent validation data set The model is verified and its performance is evaluated by comparing the consistency between the model prediction results and the actual damage conditions. This helps to identify potential problems in the model and make necessary adjustments and optimizations to ensure the accuracy and reliability of the model. The calculated comprehensive damage index of the cable structure can be used to evaluate the overall health of the cable and provide decision support for bridge operation and maintenance management. This helps bridge management departments to promptly understand the damage conditions of the cables and formulate effective maintenance and reinforcement plans to ensure the safety and durability of the bridge. This step improves the accuracy and reliability of bridge cable damage detection through multi-dimensional comprehensive analysis, high-level feature extraction, decision rule optimization, model verification and adjustment, and decision support and health assessment.

[0032] S5. Compare the comprehensive disease index of the cable structure with a preset safety threshold to generate a graded warning signal; Different levels of safety thresholds are set according to bridge design specifications, historical disease data and expert experience; the preset safety thresholds include different levels such as minor disease, moderate disease and severe disease to reflect the severity of cable disease; Compare the calculated comprehensive damage index of the cable structure with the preset safety threshold; According to the comparison results, the corresponding graded warning signals are generated; when the disease index is lower than the slight disease threshold, no warning is issued; when the disease index is between the slight disease and moderate disease thresholds, a slight warning is issued; when the disease index exceeds the moderate disease threshold but does not reach the severe disease threshold, a moderate warning is issued; when the disease index exceeds the severe disease threshold, a severe warning is issued; Deliver the generated warning signals to the bridge management department or relevant maintenance personnel in a timely manner; The graded warning signals include normal, mild warning, moderate warning, severe warning and critical warning; Normal: The cable structure comprehensive disease index is lower than the preset minimum safety threshold, indicating that the cable is in good condition and no maintenance or reinforcement measures are required; Mild warning: The comprehensive damage index of the cable structure is slightly higher than the minimum safety threshold, but still within the acceptable range. At this time, attention should be paid to the damage of the cable and regular inspections should be planned to ensure that the damage does not develop further. Moderate warning: The cable structure comprehensive disease index exceeds the mild warning threshold, but does not reach the severe warning level; this indicates that the cable has obvious disease and needs to be inspected and repaired as soon as possible to prevent the disease from worsening; Severe warning: The comprehensive damage index of the cable structure approaches or exceeds the preset severe warning threshold; at this point, the damage to the cable is already quite serious and may have a significant impact on its bearing capacity and service life; measures should be taken immediately to repair or reinforce it to ensure the safety of the bridge; Critical warning: The comprehensive disease index of the cable structure far exceeds the severe warning threshold, indicating that the cable is in an extremely dangerous state and may break or fail at any time; the emergency response mechanism should be activated immediately and emergency measures should be taken to repair or replace the cables to ensure the safety and stability of the bridge.

[0033] In this step, by setting different levels of safety thresholds and comparing the calculated cable structure comprehensive disease index with these preset thresholds, this step generates an early warning signal with a clear classification standard; it not only provides an intuitive reflection of the severity of the cable disease, but also greatly enhances the pertinence and efficiency of bridge management and maintenance; specifically, when the disease index is lower than the slight disease threshold, no early warning is issued, avoiding unnecessary interference and waste of resources; when the disease index gradually increases and reaches the slight, moderate and severe warning ranges respectively, the corresponding level of early warning signal can timely remind the bridge management department or relevant maintenance personnel to pay attention to the disease of the cable, and plan or immediately take necessary inspection and maintenance measures to effectively prevent the further development and deterioration of the disease; especially when the disease index exceeds the severe disease threshold and reaches the critical warning level, the emergency response mechanism can be immediately activated, and emergency measures can be taken to repair or replace the cable, thereby ensuring the safety and stability of the bridge and avoiding possible major safety accidents; therefore, step S5 not only improves the accuracy and timeliness of bridge cable disease detection through a scientific and reasonable graded early warning mechanism, but also effectively improves the efficiency and safety of bridge management and maintenance.

[0034] Embodiment 2: Figure 2 As shown, the bridge cable disease detection system based on image recognition of the present invention specifically includes the following modules: Data acquisition module, used to collect multi-source image data of bridge cables and environmental parameter monitoring data; A feature fusion processing module, used for performing multi-scale feature fusion processing on the multi-source image data to obtain a Lasso fusion feature tensor; An environment elimination processing module, considering the environment parameter monitoring data, performs environment elimination on each feature in the Lasso fusion feature tensor to obtain a Lasso fusion elimination feature tensor; A composite disease assessment module, used for inputting the cable fusion elimination feature tensor into a preset composite disease assessment model to obtain a comprehensive disease index of the cable structure; The early warning generation module is used to compare the comprehensive disease index of the cable structure with a preset safety threshold to generate a graded early warning signal.

[0035] The system can quickly obtain cable multi-source image data and environmental parameter monitoring data through an automated data acquisition module. Compared with manual inspection, it greatly improves the detection speed, can meet the detection needs of large-scale bridge cable networks, and solves the problem of low efficiency of existing methods. Multi-source image data includes visible light, infrared thermal imaging and 3D laser point cloud data. With the corresponding feature extraction technology, it can accurately extract the surface crack distribution, internal corrosion and cross-sectional geometric distortion of the cable, overcoming the limitations of visual inspection affected by subjective factors and poor accuracy, and can more accurately detect hidden defects. The feature fusion processing module fuses multiple features to form a cable fusion feature tensor, and then obtains the cable fusion elimination feature tensor through the environmental elimination processing module to provide comprehensive data for the composite disease assessment module; this module can conduct a comprehensive quantitative assessment of the comprehensive diseases of the cable structure through the preset composite disease assessment model, changing the situation that the existing method can only assess a single disease, and can give a timely and accurate judgment on the overall health status of the cable; It reduces the reliance on a large number of testers, reduces labor costs, and improves test efficiency. In the long run, it reduces the overall test cost and solves the problem of high cost of existing methods. The early warning generation module can compare the comprehensive disease index of the cable structure with the preset safety threshold and quickly generate a graded early warning signal, so that relevant personnel can take timely measures to avoid major safety accidents, making up for the defect that the existing method cannot provide timely early warning; In summary, the bridge cable defect detection system based on image recognition has significant advantages such as high efficiency, accuracy, comprehensiveness, intelligence and cost-effectiveness, and can effectively solve the limitations of existing bridge cable defect detection methods.

[0036] The various variations and specific embodiments of the bridge cable defect detection method based on image recognition in the aforementioned embodiment 1 are also applicable to the bridge cable defect detection system based on image recognition in the present embodiment. Through the aforementioned detailed description of the bridge cable defect detection method based on image recognition, those skilled in the art can clearly know the implementation method of the bridge cable defect detection system based on image recognition in the present embodiment, so for the sake of brevity of the specification, it will not be described in detail here.

[0037] In addition, the present application also provides an electronic device, including a bus, a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor. The transceiver, the memory, and the processor are respectively connected via a bus. When the computer program is executed by the processor, each process of the above-mentioned method for controlling output data is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be described here.

[0038] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A bridge cable disease detection method based on image recognition, characterized in that: The method comprises: Collect multi-source image data and environmental parameter monitoring data of bridge cables; Performing multi-scale feature fusion processing on the multi-source image data to obtain a Lasso fusion feature tensor; Considering the environmental parameter monitoring data, performing environmental elimination on each feature in the Lasso fusion feature tensor to obtain a Lasso fusion elimination feature tensor; Inputting the cable fusion elimination characteristic tensor into a preset composite disease assessment model to obtain a comprehensive disease index of the cable structure; The comprehensive disease index of the cable structure is compared with a preset safety threshold to generate a graded early warning signal.

2. The bridge cable disease detection method based on image recognition according to claim 1 is characterized in that: The multi-source image data includes visible light band image data, infrared thermal imaging data and three-dimensional laser point cloud data.

3. The bridge cable disease detection method based on image recognition according to claim 1 is characterized in that: The cable fusion feature tensor includes surface crack distribution features, cable internal corrosion features and cross-sectional geometric distortion features; wherein the surface crack distribution features are extracted based on visible light band image data through a spatial domain neural network model; the cable internal corrosion features are analyzed based on infrared thermal imaging data through a frequency domain transformation algorithm; the cross-sectional geometric distortion features are obtained based on three-dimensional laser point cloud data through three-dimensional point cloud matching calculation.

4. The bridge cable disease detection method based on image recognition according to claim 3 is characterized in that: The process of acquiring the surface crack distribution characteristics includes: Preprocessing the visible light band image data, wherein the preprocessing method includes image denoising, image enhancement and image segmentation; Collect visible light image data containing cracks on the cable surface and build a spatial domain neural network model; The preprocessed visible light band image is input into the spatial domain neural network model to obtain the surface crack distribution characteristics.

5. The bridge cable disease detection method based on image recognition according to claim 3 is characterized in that: The process of obtaining the internal corrosion characteristics of the cable includes: Preprocessing the infrared thermal imaging data, wherein the preprocessing steps include image enhancement, denoising, and temperature calibration and normalization; The pre-processed infrared thermal imaging data is converted from the time domain to the frequency domain using the fast Fourier transform algorithm; Perform filtering on the frequency domain data after fast Fourier transformation; The internal corrosion characteristics of the cable are extracted from the filtered frequency domain data.

6. The bridge cable disease detection method based on image recognition according to claim 3 is characterized in that: The process of acquiring the cross-sectional geometric distortion feature includes: Use a 3D laser scanner to scan the bridge cables and obtain 3D laser point cloud data of the cables; Preprocessing the collected original 3D laser point cloud data; the preprocessing steps include removing noise points, normalizing data, and registering point clouds; According to the spatial position and shape of the cable, the cross-sectional point cloud of the cable is extracted from the pre-processed three-dimensional point cloud data; Matching and calculating the extracted cable section point cloud with the preset cable standard section model; According to the matching results, the geometric distortion characteristics of the cable section are calculated.

7. The bridge cable disease detection method based on image recognition according to claim 1, characterized in that: The graded warning signals include normal, mild warning, moderate warning, severe warning and critical warning.

8. A bridge cable disease detection system based on image recognition, characterized in that: The system comprises: Data acquisition module, used to collect multi-source image data of bridge cables and environmental parameter monitoring data; A feature fusion processing module, used for performing multi-scale feature fusion processing on the multi-source image data to obtain a Lasso fusion feature tensor; An environment elimination processing module, considering the environment parameter monitoring data, performs environment elimination on each feature in the Lasso fusion feature tensor to obtain a Lasso fusion elimination feature tensor; A composite disease assessment module, used for inputting the cable fusion elimination feature tensor into a preset composite disease assessment model to obtain a comprehensive disease index of the cable structure; The early warning generation module is used to compare the comprehensive disease index of the cable structure with a preset safety threshold to generate a graded early warning signal.

9. An electronic device, comprising a bus, a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the transceiver, the memory, and the processor are connected via the bus, wherein: When the computer program is executed by the processor, the steps in the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps in the method according to any one of claims 1 to 7 are implemented.

Images