Visualization-based power transmission detection system

Abstract

The invention, which relates to the field of power transmission detection, discloses a visualization-based power transmission detection system comprising an image acquisition module, a preprocessing module, a feature extraction module, a defect identification module and an alarm module. A high-definition camera is used for collecting images, a path is planned according to an inspection task during preparation, equipment is debugged, shake is prevented in the process, and a continuous shooting mode is selected; noise reduction is carried out on the image, the image is enhanced after being processed through a wavelet transform algorithm, and the image is standardized into a specific resolution ratio and a specific pixel value through self-adaptive histogram equalization operation; extracting multi-scale features based on a deep learning model, extracting frequency domain features during defect identification, fusing the frequency domain features into vectors, inputting the vectors into a classifier, and optimizing through a feature pyramid network; and finally, sound-light alarm is generated when defects are detected, information is uploaded to a remote monitoring center, geographic information system linkage and local storage are supported, data distribution and decision support are ensured to be timely and accurate, and power transmission equipment is effectively maintained.

Description

Technical Field

[0001] The present invention relates to the technical field of power transmission detection, and more particularly, to a visualization-based power transmission detection system. Background Art

[0002] With the continuous expansion of power grids and increasing demands for power supply reliability, the maintenance and management of transmission equipment has become a crucial component of power system operations. Over the long term, transmission lines are exposed to a variety of factors, including environmental stress, equipment aging, and external shocks. These factors can lead to defects such as insulator cracks, conductor corrosion, and dangling foreign objects, impacting the safety and stability of power transmission.

[0003] However, in practice, this approach still has some shortcomings. For example, traditional power transmission inspection methods rely primarily on manual inspections, which have certain limitations. This method is not only time-consuming and labor-intensive, but also hinders comprehensive and systematic equipment inspections due to human factors. In complex environments, manual inspections are difficult to ensure in terms of accuracy and timeliness, potentially leading to missed detections or misjudgments of potential defects. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a visualization-based power transmission detection system to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] Image acquisition module: used to acquire high-resolution images of transmission lines using a high-resolution camera;

[0007] Preprocessing module: used for performing noise reduction, enhancement and standardization processing on the image or video data;

[0008] Feature extraction module: used to extract multi-scale features from pre-processed images based on a deep learning model, including structural features and surface abnormality features of power transmission equipment;

[0009] Defect identification module: used to input the extracted features into the classifier to identify the type and location of defects in the power transmission equipment, such as broken insulators, rusted conductors, or hanging foreign objects;

[0010] Alarm module: used to generate an alarm signal and output defect location information when a defect is detected.

[0011] Preferably, in the image acquisition module, preliminary preparations are made and a high-definition camera and lens that supports 4K resolution are selected, including a telephoto lens and a wide-angle lens; the telephoto lens is used for long-distance shooting, and the wide-angle lens is used for large-area coverage; and a tripod stabilization device is equipped to prevent shooting shake; at the same time, the lighting conditions are checked to ensure sufficient light during shooting, LED fill light equipment is used, and shooting is avoided in severe weather such as strong wind, rain and snow; according to the inspection task, the drone flight path or the installation location of the fixed equipment is planned in advance to ensure coverage of all target areas.

[0012] Preferably, in the pre-processing module, during the image processing, noise reduction is first performed. A noise reduction algorithm based on wavelet transform is used to first perform multi-scale wavelet decomposition on the input image to obtain high-frequency components and low-frequency components. Then, soft threshold processing is performed on the high-frequency components to remove noise components. Then, wavelet reconstruction is performed on the processed high-frequency components and low-frequency components to obtain a noise-reduced image.

[0013] The specific method of wavelet transform is:

[0014] f(c, d) = ∑ h ∑ k ∑ l c h,k,l Ψ h,k,l (c,d), where c h,k,l is represented by the wavelet coefficient, h is represented by the scale parameter, Ψ h,k,l (c, d) is expressed as a wavelet function, which represents the wavelet basis function at scale h and position (k, l);

[0015] After noise reduction, image enhancement is performed. Through adaptive histogram equalization and contrast-limited adaptive histogram equalization, the image is first divided into several local areas, and then histogram equalization is performed on each local area to improve the local contrast. Finally, bilinear interpolation is performed on the equalized areas to eliminate blocking effects.

[0016] The specific calculation method of the histogram is:

[0017] Among them, P(f) is expressed as a histogram, n f It is expressed as the number of pixels of gray level f, and N is the total number of pixels in the image;

[0018] The calculation method of the cumulative distribution function is as follows:

[0019] Where R(f) represents the cumulative distribution function, P(f) represents the histogram, and f represents the grayscale;

[0020] After obtaining the new grayscale, the equalized image is generated by replacing the value of each grayscale in the image. Assuming that the grayscale value of a pixel in the original image is f, the new grayscale value after equalization is P″(f). That is, the value of each pixel in the equalized image can be obtained through mapping.

[0021] The calculation method of the equalized image is as follows:

[0022] S(c, d) = P″(m(c, d)), where S(c, d) represents the equalized image and m(c, d) represents the pixel value of the original image;

[0023] Finally, the enhanced image is normalized to a preset resolution of 224×224 pixels, and its pixel values are normalized to a mean of 0 and a standard deviation of 1.

[0024] Preferably, in the feature extraction module, the structural features include the geometric shape, size, position and relative relationship of the power transmission equipment; the structural features are extracted by a convolutional neural network in a deep learning model, and the calculation method of edge extraction is specifically as follows:

[0025] Where E(x, y) represents the edge strength at position (x, y), G x Expressed as the gradient in the x direction, G y Expressed as the gradient in the y direction;

[0026] The calculation method of the gray level co-occurrence matrix is as follows:

[0027] Where P(i, j) represents the co-occurrence frequency between grayscale values i and j, N represents the total number of pixels in the image used to calculate the co-occurrence matrix, I(a, b) represents the grayscale value of the input image at position (a, b), and d a It is expressed as the relative offset in the a direction, a represents the horizontal axis direction, d b It is expressed as the relative offset in the b direction, b represents the longitudinal axis direction, and δ represents the Kronecker delta function;

[0028] The calculation method of the local binary pattern value is as follows:

[0029] Where LBP(x,y) represents the local binary pattern value at position (x,y), p represents the total number of neighborhood pixels, g(p) represents the grayscale value of the neighboring pixel p, g(c) represents the grayscale value of the center pixel c, and s represents the sign function;

[0030] The structural features of the power transmission equipment are obtained by edge extraction, gray-level co-occurrence matrix and local binary pattern value;

[0031] The surface abnormality features are extracted through the Feature Pyramid Network (FPN), a multi-scale feature pyramid is constructed, and feature maps of different resolutions are fused; the fused feature maps are upsampled to generate high-resolution abnormality feature maps.

[0032] Preferably, in the defect recognition module, the frequency domain features of the image are extracted by spectrum analysis technology, the abnormal frequency distribution of the defect area is captured, and the input image is subjected to fast Fourier transform (FFT) to convert the image from the spatial domain to the frequency domain to generate a spectrum diagram; then, the frequency features of the defect area, the high-frequency components of the insulator rupture area and the low-frequency abnormal distribution of the conductor corrosion area are extracted by high-pass filtering technology; the extracted frequency domain features are fused to form a multimodal feature vector, which is input into a multimodal classifier; the classifier adopts a hybrid architecture combining deep learning and classical machine learning, wherein the target detection module based on YOLOv8 decodes the multimodal features, extracts high-level semantic information through convolutional layers and fully connected layers, and outputs the defect type, which includes insulator rupture, conductor corrosion or foreign object hanging and its position coordinates in the image, and the position coordinates are represented by a bounding box; at the same time, for structural features, a random forest classifier is used for auxiliary decision-making, key features are screened by feature importance evaluation, and combined with the output results of the deep learning model, a weighted average mechanism is used to improve the classification accuracy, especially in small sample or noise interference scenarios. The robustness of the model is significantly improved.

[0033] Preferably, in the alarm module, when the defect recognition module detects a defect such as insulator cracking, conductor corrosion or hanging foreign objects, the defect type, location coordinates and confidence information are transmitted to the alarm module; then, whether to trigger an alarm is determined based on a preset defect severity threshold; if the alarm conditions are met, the module generates an alarm signal in the following manner: at the local device end, triggering an audible and visual alarm device, wherein the alarm device is a buzzer and an LED indicator light, and generating an alarm log containing the defect type, location coordinates, detection time and device ID; at the same time, uploading the alarm information to the remote monitoring center in real time via wireless communication.

[0034] The technical effects and advantages of the present invention are as follows:

[0035] The present invention uses high-definition cameras and compatible lenses that support 4K resolution, plans routes and debugs equipment according to inspection tasks, and ensures stable and clear shooting. During image preprocessing, noise reduction is first performed based on wavelet transform, and then enhanced through adaptive histogram equalization, and finally standardized to a specific size and pixel value. Feature extraction uses a deep learning model, covers structural and surface abnormality features, and uses a variety of calculation methods. Defect recognition uses spectrum analysis technology to extract frequency domain features, fuses multimodal feature vectors through a classifier, and combines with a random forest classifier to improve accuracy and robustness. It also uses super-resolution technology to enhance the visibility and detection accuracy of small defects. The alarm module triggers an audible and visual alarm when a defect is detected and uploads information to a remote monitoring center. The present invention uses advanced image recognition technology and machine learning algorithms, and the visualization system can accurately identify various small defects. For minor damage to insulators, small cracks on conductors, etc., the system can make accurate judgments based on pre-trained models; and can improve the accuracy and timeliness of detection in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a schematic diagram of module connection of the present invention. DETAILED DESCRIPTION

[0037] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0038] See also Figure 1 As shown, the present invention provides a visualization-based power transmission detection system, which includes an image acquisition module, a preprocessing module, a feature extraction module, a defect recognition module and an alarm module.

[0039] Image acquisition module: used to acquire high-resolution images of transmission lines using a high-resolution camera;

[0040] In the image acquisition module, preliminary preparations are made to select a high-definition camera and lens that supports 4K resolution, including a telephoto lens and a wide-angle lens; the telephoto lens is used for long-distance shooting, and the wide-angle lens is used for large-area coverage; and a tripod stabilization device is used to prevent shooting shake. At the same time, the lighting conditions are checked to ensure sufficient lighting during shooting, LED fill lighting equipment is used, and shooting in inclement weather such as strong winds, rain, and snow is avoided. According to the inspection task, the flight path of the drone or the installation location of the fixed equipment is planned in advance to ensure coverage of all target areas.

[0041] Next, debug the equipment, set the camera resolution to 4K, adjust the shutter speed, aperture, and ISO parameters, and select the autofocus mode; calibrate the tripod to ensure there is no shaking during shooting, and test the effectiveness of the fill light equipment in low light conditions. During the image acquisition stage, aim the camera at the power transmission line, adjust the angle and height, and select the continuous shooting mode for shooting; during the shooting process, check the image or video quality in real time.

[0042] Preprocessing module: used for performing noise reduction, enhancement and standardization processing on the image or video data;

[0043] In the pre-processing module, during the image processing, noise reduction is first performed. A noise reduction algorithm based on wavelet transform is used to first perform multi-scale wavelet decomposition on the input image to obtain high-frequency and low-frequency components. Then, soft threshold processing is performed on the high-frequency components to remove noise components. The processed high-frequency and low-frequency components are then reconstructed using wavelet transform to obtain a denoised image.

[0044] The specific method of wavelet transform is:

[0045] f(c, d) = ∑ h ∑ k ∑ l c h,k,l Ψ h,k,l (c,d), where c h,k,l is represented by the wavelet coefficient, h is represented by the scale parameter, Ψ h,k,l (c, d) is expressed as a wavelet function, which represents the wavelet basis function at scale h and position (k, l);

[0046] After noise reduction, image enhancement is performed. Through adaptive histogram equalization and contrast-limited adaptive histogram equalization, the image is first divided into several local areas, and then histogram equalization is performed on each local area to improve the local contrast. Finally, bilinear interpolation is performed on the equalized areas to eliminate blocking effects.

[0047] The specific calculation method of the histogram is:

[0048] Among them, P(f) is expressed as a histogram, n f It is expressed as the number of pixels of gray level f, and N is the total number of pixels in the image;

[0049] The calculation method of the cumulative distribution function is as follows:

[0050] Where R(f) represents the cumulative distribution function, P(f) represents the histogram, and f represents the grayscale;

[0051] After obtaining the new grayscale, the equalized image is generated by replacing the value of each grayscale in the image. Assuming that the grayscale value of a pixel in the original image is f, the new grayscale value after equalization is P″(f). That is, the value of each pixel in the equalized image can be obtained through mapping.

[0052] The calculation method of the equalized image is as follows:

[0053] S(c, d) = P″(m(c, d)), where S(c, d) represents the equalized image and m(c, d) represents the pixel value of the original image;

[0054] Finally, the enhanced image is normalized to a preset resolution of 224×224 pixels, and its pixel values are normalized to a mean of 0 and a standard deviation of 1.

[0055] Feature extraction module: used to extract multi-scale features from pre-processed images based on a deep learning model, including structural features and surface abnormality features of power transmission equipment;

[0056] In the feature extraction module, structural features include the geometric shape, size, position, and relative relationship of the power transmission equipment. Structural features are extracted through a convolutional neural network in a deep learning model. The edge extraction calculation method is specifically as follows:

[0057] Where E(x, y) represents the edge strength at position (x, y), G x Expressed as the gradient in the x direction, G y Expressed as the gradient in the y direction;

[0058] The calculation method of the gray level co-occurrence matrix is as follows:

[0059] Where P(i, j) represents the co-occurrence frequency between grayscale values i and j, N represents the total number of pixels in the image used to calculate the co-occurrence matrix, I(a, b) represents the grayscale value of the input image at position (a, b), and d a It is expressed as the relative offset in the a direction, a represents the horizontal axis direction, d b It is expressed as the relative offset in the b direction, b represents the longitudinal axis direction, and δ represents the Kronecker delta function;

[0060] The calculation method of the local binary pattern value is as follows:

[0061] Where LBP(x,y) represents the local binary pattern value at position (x,y), p represents the total number of neighborhood pixels, g(p) represents the grayscale value of the neighboring pixel p, g(c) represents the grayscale value of the center pixel c, and s represents the sign function;

[0062] The structural features of the power transmission equipment are obtained by edge extraction, gray-level co-occurrence matrix and local binary pattern value;

[0063] The surface abnormality features are extracted through the Feature Pyramid Network (FPN), a multi-scale feature pyramid is constructed, and feature maps of different resolutions are fused; the fused feature maps are upsampled to generate high-resolution abnormality feature maps.

[0064] Defect identification module: used to input the extracted features into the classifier to identify the type and location of defects in the power transmission equipment, such as broken insulators, rusted conductors, or hanging foreign objects;

[0065] In the defect recognition module, frequency domain features of the image are extracted through spectrum analysis technology to capture the abnormal frequency distribution of the defect area. A fast Fourier transform (FFT) is performed on the input image to convert the image from the spatial domain to the frequency domain to generate a spectrum diagram. Next, high-pass filtering technology is used to extract the frequency characteristics of the defect area, the high-frequency components of the insulator crack area and the low-frequency abnormal distribution of the conductor corrosion area. The extracted frequency domain features are fused to form a multimodal feature vector, which is input into the multimodal classifier. The classifier adopts a hybrid architecture combining deep learning and classical machine learning. The target detection module based on YOLOv8 decodes the multimodal features, extracts high-level semantic information through convolutional layers and fully connected layers, and outputs the defect type, which includes insulator cracks, conductor corrosion, or foreign objects hanging, and their position coordinates in the image. The position coordinates are represented by bounding boxes. At the same time, a random forest classifier is used to assist in decision-making for structural features. Key features are selected through feature importance evaluation. Combined with the output results of the deep learning model, a weighted average mechanism is used to improve classification accuracy, especially in scenarios with small samples or noise interference. The robustness of the model is significantly improved.

[0066] For complex scenarios, the module fuses multi-scale features through a feature pyramid network to enhance the detection capability of wire corrosion points: FPN combines low-level detail features with high-level semantic features through top-down and bottom-up feature fusion paths to ensure that defective areas can be accurately located at different scales. In addition, the module combines super-resolution technology to enhance defective areas in low-quality images and reconstructs high-resolution images through generative adversarial networks, significantly improving the visibility and detection accuracy of small defects. Finally, the module optimizes the detection results through an improved non-maximum suppression algorithm to avoid missed detection or false detection problems in dense areas; at the same time, the positioning results are spatially constrained and filtered based on the spatial distribution pattern of the insulator strings of the transmission equipment structure, and abnormal detection frames that do not conform to the equipment structure characteristics are eliminated through geometric rules to further ensure the accuracy and reliability of the recognition results; through the above-mentioned multi-dimensional feature extraction, multimodal fusion and post-processing optimization, the defect recognition module can achieve high-precision and high-robustness defect detection and positioning in complex environments.

[0067] Alarm module: used to generate an alarm signal and output defect location information when a defect is detected;

[0068] In the alarm module, when the defect recognition module detects a cracked insulator, corroded conductor, or a foreign object hanging defect, the defect type, location coordinates, and confidence level information are transmitted to the alarm module; then, based on a preset defect severity threshold, the module determines whether to trigger an alarm; if the alarm condition is met, the module generates an alarm signal in the following manner: on the local device side, triggering an audible and visual alarm device, which includes a buzzer and an LED indicator light, and generating an alarm log containing the defect type, location coordinates, detection time, and device ID; and simultaneously uploading the alarm information to a remote monitoring center in real time via wireless communication;

[0069] The alarm information is encapsulated in a structured JSON format, including the defect type, location coordinates, confidence level, and timestamp. The module then supports linkage with the geographic information system, mapping the defect location coordinates to a map interface, and automatically generating inspection work orders, which are then pushed to relevant maintenance personnel. To cope with network anomalies, the module has a built-in local storage function that caches alarm information on the device side and re-uploads it after the network is restored to ensure data is not lost. The alarm module achieves real-time distribution of alarm information through multi-channel collaboration, providing timely and accurate decision-making support for transmission equipment maintenance.

[0070] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A power transmission detection system based on visualization, characterized in that: include: Image acquisition module: used to acquire high-resolution images of transmission lines using a high-resolution camera; Preprocessing module: used for performing noise reduction, enhancement and standardization processing on the image or video data; Feature extraction module: used to extract multi-scale features from pre-processed images based on a deep learning model, including structural features and surface abnormality features of power transmission equipment; Defect identification module: used to input the extracted features into the classifier to identify the type and location of defects in the power transmission equipment, such as broken insulators, rusted conductors, or hanging foreign objects; Alarm module: used to generate an alarm signal and output defect location information when a defect is detected.

2. The visualization-based power transmission detection system according to claim 1, characterized in that: In the image acquisition module, preliminary preparations are made and a high-definition camera and lens that supports 4K resolution are selected. The lenses include a telephoto lens and a wide-angle lens; the telephoto lens is used for long-distance shooting, and the wide-angle lens is used for large-area coverage; and a tripod stabilization device is equipped to prevent shooting shake; at the same time, the lighting conditions are checked to ensure sufficient light during shooting, LED fill lighting equipment is used, and shooting in bad weather such as strong winds, rain and snow is avoided; according to the inspection task, the drone flight path or the installation location of the fixed equipment is planned in advance to ensure coverage of all target areas.

3. The visualization-based power transmission detection system according to claim 1, characterized in that: In the preprocessing module, during the image processing, noise reduction is first performed; a noise reduction algorithm based on wavelet transform is used to first perform multi-scale wavelet decomposition on the input image to obtain high-frequency components and low-frequency components, then soft threshold processing is performed on the high-frequency components to remove noise components, and then wavelet reconstruction is performed on the processed high-frequency components and low-frequency components to obtain a noise-reduced image; the wavelet transform method is specifically as follows: f(c, d) = ∑ h ∑ k ∑ l c h,k,l Ψ h,k,l (c,d), where c h,k,l is represented by the wavelet coefficient, h is represented by the scale parameter, Ψ h,k,l (c, d) is expressed as a wavelet function, which represents the wavelet basis function at scale h and position (k, l).

4. The visualization-based power transmission detection system according to claim 3, characterized in that: After noise reduction, image enhancement is performed. Through adaptive histogram equalization and contrast-limited adaptive histogram equalization, the image is first divided into several local areas, and then histogram equalization is performed on each local area to improve the local contrast. Finally, bilinear interpolation is performed on the equalized areas to eliminate blocking effects. The specific calculation method of the histogram is: Among them, P(f) is expressed as a histogram, n f It is expressed as the number of pixels of gray level f, and N is the total number of pixels in the image; The calculation method of the cumulative distribution function is as follows: Where R(f) represents the cumulative distribution function, P(f) represents the histogram, and f represents the grayscale; After obtaining the new grayscale, the equalized image is generated by replacing the value of each grayscale in the image. Assuming that the grayscale value of a pixel in the original image is f, the new grayscale value after equalization is P″(f). That is, the value of each pixel in the equalized image can be obtained through mapping. The calculation method of the equalized image is as follows: S(c, d) = P″(m(c, d)), where S(c, d) represents the equalized image and m(c, d) represents the pixel value of the original image; Finally, a standardization process is performed to adjust the size of the enhanced image to the preset resolution, which is 224×224 pixels. At the same time, its pixel values are normalized so that their mean is 0 and their standard deviation is 1.

5. The visualization-based power transmission detection system according to claim 1, characterized in that: In the feature extraction module, the structural features include the geometric shape, size, position and relative relationship of the power transmission equipment; Structural features are extracted through the convolutional neural network in the deep learning model. The calculation method for edge extraction is as follows: Where E(x, y) represents the edge strength at position (x, y), G x Expressed as the gradient in the x direction, G y Expressed as the gradient in the y direction.

6. The visualization-based power transmission detection system according to claim 1, characterized in that: The calculation method of the gray level co-occurrence matrix is as follows: Where P(i, j) represents the co-occurrence frequency between grayscale values i and j, N represents the total number of pixels in the image used to calculate the co-occurrence matrix, I(a, b) represents the grayscale value of the input image at position (a, b), and d a It is expressed as the relative offset in the a direction, a represents the horizontal axis direction, d b It is expressed as the relative offset in the b direction, b represents the longitudinal axis direction, and δ represents the Kronecker delta function; The calculation method of the local binary pattern value is as follows: Where LBP(x,y) represents the local binary pattern value at position (x,y), p represents the total number of neighborhood pixels, g(p) represents the grayscale value of the neighboring pixel p, g(c) represents the grayscale value of the center pixel c, and s represents the sign function; The structural features of the transmission equipment are obtained through edge extraction, gray-level co-occurrence matrix and local binary pattern value.

7. The visualization-based power transmission detection system according to claim 1, characterized in that: In the defect recognition module, frequency domain features of the image are extracted using spectrum analysis technology to capture the abnormal frequency distribution of the defect area. A fast Fourier transform (FFT) is performed on the input image to convert the image from the spatial domain to the frequency domain, generating a spectrum graph. Next, high-pass filtering is used to extract the frequency characteristics of the defect area, including the high-frequency components of the insulator crack area and the low-frequency abnormal distribution of the conductor corrosion area. The extracted frequency domain features are fused to form a multimodal feature vector, which is input into a multimodal classifier. The classifier adopts a hybrid architecture that combines deep learning and classical machine learning. The YOLOv8-based target detection module decodes the multimodal features, extracts high-level semantic information through convolutional layers and fully connected layers, and outputs the defect type, including insulator crack, conductor corrosion, or foreign object hanging, and its location coordinates in the image, represented by a bounding box. At the same time, a random forest classifier is used to assist in decision-making for structural features. Key features are selected through feature importance evaluation. Combined with the output of the deep learning model, a weighted averaging mechanism is used to improve classification accuracy, significantly improving the robustness of the model in scenarios with small samples or noise interference.

8. The visualization-based power transmission detection system according to claim 1, characterized in that: In the alarm module, when the defect recognition module detects insulator cracks, conductor corrosion or foreign object hanging defects, the defect type, location coordinates and confidence information are transmitted to the alarm module; then, based on the preset defect severity threshold, it is determined whether to trigger an alarm; if the alarm conditions are met, the module generates an alarm signal in the following manner: at the local device end, an audible and visual alarm device is triggered, wherein the alarm device is a buzzer and an LED indicator light, and an alarm log containing the defect type, location coordinates, detection time and device ID is generated; at the same time, the alarm information is uploaded to the remote monitoring center in real time via wireless communication.

Images