Power transmission line insulator defect detection system based on multispectral imaging

Abstract

The invention provides a power transmission line insulator defect detection system based on multispectral imaging, and relates to the technical field of defect detection, and the system comprises the steps: determining different levels of defect feature channels in a defect image of an insulator through all insulator defect feature information; determining a defect evaluation index of the defect position of the power transmission line insulator according to the defect associated information, and positioning the surface defect position of the power transmission line insulator according to the defect evaluation index; determining the defect prediction deviation of the defect position of the power transmission line insulator according to the size of the anchor frame and the defect feature channel, and layering the surface defect state of the power transmission line insulator according to the defect prediction deviation; and determining a defect comparison unit on the surface of the insulator during defect detection of the power transmission line insulator, and further detecting the defect of the power transmission line insulator according to the defect comparison unit. According to the method, the comprehensiveness of defect detection of the insulator of the power transmission line can be improved under the influence of deviation of defect positioning and evaluation results.

Description

Technical Field

[0001] The present application relates to the technical field of defect detection, and more specifically, to a transmission line insulator defect detection system based on multi-spectral imaging. Background Art

[0002] Defect detection refers to the process of identifying, locating and evaluating abnormal conditions or damage characteristics on the surface or inside of a target object through technical means. In the detection of transmission line insulators, defect detection mainly targets problems such as cracks, aging, contamination, and peeling on the surface of the insulators. These defects may lead to reduced insulation performance and even safety hazards. Defect detection technology based on multispectral imaging captures its reflection characteristics and material differences in different bands by collecting multispectral images of the insulator surface, thereby revealing defect information that is invisible to the human eye. Multispectral imaging can perform non-contact detection of insulators, is not affected by light or environmental changes, and is suitable for complex outdoor environments. Combined with intelligent algorithms, through image segmentation, feature extraction and classification models, it is possible to achieve accurate positioning and hierarchical evaluation of defects, providing efficient and reliable technical support for condition monitoring and preventive maintenance of transmission lines. This method significantly improves detection efficiency and accuracy and reduces the limitations of manual inspections.

[0003] Existing transmission line insulator defect detection systems based on multispectral imaging usually use multispectral imagers installed on transmission line inspection equipment to collect multi-band images of the insulator surface in real time. These images cover the visible light to infrared bands, capturing subtle differences in the insulator surface, such as cracks, stains, aging or corrosion defects. After image acquisition, the system uses image processing and analysis algorithms, such as image preprocessing, feature extraction and image segmentation, to identify potential defect areas. Through machine learning or deep learning models, the system can automatically determine the type and severity of defects based on the extracted spectral features, and classify and locate the defects according to preset standards. The detection results are then presented through a visual interface, providing the defect location, type and risk assessment. However, in the transmission line insulator defect detection based on multispectral imaging, there are defects in data acquisition and processing accuracy, that is, the settings of the multispectral imager are not uniform, the image segmentation is not accurate, and the sensitivity of defect recognition is insufficient, making it difficult for the system to stably capture all types of defect features in a complex environment, resulting in deviations in defect location and evaluation results, and thus failing to effectively ensure the comprehensiveness and accuracy of the detection. Therefore, how to improve the comprehensiveness of transmission line insulator defect detection under the influence of deviations in defect location and evaluation results is a problem faced by the industry. Summary of the invention

[0004] The present application provides a transmission line insulator defect detection system based on multi-spectral imaging, which can improve the comprehensiveness of transmission line insulator defect detection under the influence of deviations in defect positioning and evaluation results.

[0005] The present application provides a transmission line insulator defect detection system based on multispectral imaging, the defect detection system comprising:

[0006] An image acquisition module is used to obtain a surface multispectral image of the transmission line insulator by setting the same data acquisition range for each multispectral imager and performing multispectral imaging data acquisition on the surface of the insulator to be detected on the transmission line;

[0007] A defect feature determination module is used to perform mutual projection segmentation on the surface multi-spectral image to obtain multiple types of insulator defect feature information, and determine defect feature channels of different levels in the defect image of the insulator through all the insulator defect feature information;

[0008] A defect location positioning module, used to obtain defect association information on the surface of the transmission line insulator, determine a defect assessment index of the defect position of the transmission line insulator according to the defect association information, and then locate the surface defect position of the transmission line insulator according to the defect assessment index;

[0009] A defect state stratification module is used to set the anchor frame size of multispectral imaging during transmission line insulator defect detection, determine the defect prediction deviation of the transmission line insulator defect position according to the anchor frame size and the defect feature channel, and stratify the surface defect state of the transmission line insulator according to the defect prediction deviation;

[0010] The defect comparison detection module is used to determine the defect comparison unit on the insulator surface during transmission line insulator defect detection according to the surface defect position after positioning and the surface defect state after stratification, and then detect the transmission line insulator defects based on the defect comparison unit.

[0011] In this embodiment, the surface multispectral image refers to an image set formed by the reflection or emission spectrum information of the object surface at multiple specific wavelengths acquired by the multispectral imaging technology.

[0012] In this embodiment, the surface multi-spectral image is segmented and the defect feature information of multiple types of insulators is obtained, which specifically includes:

[0013] Extract specific bands from multispectral images of transmission line insulator surfaces;

[0014] Determine the region of interest when interactively mapping the multispectral image according to the specific band;

[0015] The surface multispectral image is segmented into multiple types of insulator defect feature information through the region of interest.

[0016] In this embodiment, the defect feature information refers to a specific information set reflecting the defect type and nature in the image region extracted by segmentation.

[0017] In this embodiment, the defect feature channel refers to a channel extracted according to the hierarchical division of defect features in the multispectral image.

[0018] In this embodiment, the defect correlation information refers to comprehensive information obtained by analyzing the spatial relationship, spectral characteristic correlation and mutual influence between different defect regions on the surface of the transmission line insulator.

[0019] In this embodiment, the surface defect position refers to the specific spatial position of a defect (such as crack, aging, pollution, etc.) on the surface of the transmission line insulator relative to the geometric center of the insulator or the reference coordinate system.

[0020] In this embodiment, the defect prediction deviation of determining the defect position of the transmission line insulator according to the anchor frame size and the defect characteristic channel specifically includes:

[0021] Determining defect test information of the transmission line insulator by means of the anchor frame size;

[0022] Determine the overlap sensitivity of the defect position of the transmission line insulator according to the defect characteristic channel;

[0023] The defect prediction deviation of the defect position of the transmission line insulator is determined according to the defect test information and the overlap sensitivity.

[0024] In this embodiment, the surface defect state refers to the classification result of the current physical characteristics and degradation degree of the surface defects of the transmission line insulator.

[0025] In this embodiment, the defect comparison unit refers to a specific area divided for achieving accurate comparison and evaluation during the transmission line insulator defect detection process.

[0026] The technical solution provided by the embodiments disclosed in this application has the following beneficial effects:

[0027] By setting the same data acquisition range for each multispectral imager and performing multispectral imaging data acquisition on the surface of the insulator to be detected on the transmission line, a surface multispectral image of the transmission line insulator is obtained; the surface multispectral image is segmented by mutual projection to obtain multiple types of insulator defect feature information, and defect feature channels at different levels in the defect image of the insulator are determined through all the insulator defect feature information; defect association information on the surface of the transmission line insulator is obtained, and a defect assessment index of the defect position of the transmission line insulator is determined according to the defect association information, and then the surface defect position of the transmission line insulator is located by the defect assessment index; the anchor frame size of the multispectral imaging during the transmission line insulator defect detection is set, the defect prediction deviation of the transmission line insulator defect position is determined according to the anchor frame size and the defect feature channel, and the surface defect state of the transmission line insulator is stratified according to the defect prediction deviation; the defect comparison unit of the insulator surface during the transmission line insulator defect detection is determined according to the located surface defect position and the stratified surface defect state, and then the transmission line insulator defect is detected according to the defect comparison unit.

[0028] It can be seen that in the present application, it is possible to detect defects in insulators of transmission lines under the influence of deviations in defect positioning and evaluation results; wherein, by setting the same data acquisition range and performing multispectral image acquisition under the same standard, it is possible to ensure that consistent and high-quality image data is obtained, which helps to reduce the deviation caused by inconsistent imager configuration and improve the stability and reliability of subsequent defect detection results; by accurately extracting multiple types of defect feature information on the surface of the insulator through the mutual reflection segmentation technology, it is possible to more comprehensively identify different types of defects, thereby improving the defect classification and recognition accuracy of the system, and avoiding the problems of insufficient image segmentation accuracy and inaccurate defect extraction in the prior art; by acquiring and analyzing defect association information, it is possible to classify and identify defects based on the actual location and characteristics of the defects. The feature generates an accurate defect assessment index, making defect assessment more scientific and reliable, thereby improving the accuracy of defect positioning and overcoming the technical defect of low assessment accuracy in the prior art; setting a reasonable anchor frame size and optimizing the prediction deviation in combination with the defect feature channel can effectively improve the accuracy of defect detection, especially when the defect size and position are difficult to determine, thereby improving the problem of large deviation in the prior art and improving the accuracy of defect prediction; by setting a defect comparison unit based on the surface defect position after positioning and the defect state after stratification, the surface defects of the insulator can be accurately compared and detected one by one, improving the hierarchical accuracy and automation level of defect detection, and further overcoming the problem of lack of accurate comparison and intelligent stratification in the existing detection system.

[0029] In summary, the technical solution adopted in the present application can improve the comprehensiveness of transmission line insulator defect detection under the influence of deviations in defect location and evaluation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0031] Figure 1 It is a module structure diagram of a transmission line insulator defect detection system based on multispectral imaging provided by the present application;

[0032] Figure 2 It is a schematic diagram of a process for determining a defect characteristic channel provided by the present application;

[0033] Figure 3 It is a schematic diagram of a process for determining a defect assessment index provided in this application. DETAILED DESCRIPTION

[0034] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0035] The embodiment of the present application provides a transmission line insulator defect detection system based on multispectral imaging, the core of which is to obtain a surface multispectral image of the transmission line insulator by setting the same data acquisition range for each multispectral imager and performing multispectral imaging data acquisition on the surface of the insulator to be detected on the transmission line; performing mutual projection segmentation on the surface multispectral image to obtain multiple types of insulator defect feature information, and determining defect feature channels at different levels in the defect image of the insulator through all the insulator defect feature information; obtaining defect correlation information on the surface of the transmission line insulator, and determining the transmission line insulator defect according to the defect correlation information. The defect evaluation index of the position is determined, and the surface defect position of the transmission line insulator is located by the defect evaluation index; the anchor frame size of the multispectral imaging during the transmission line insulator defect detection is set, and the defect prediction deviation of the transmission line insulator defect position is determined according to the anchor frame size and the defect feature channel, and the surface defect state of the transmission line insulator is layered according to the defect prediction deviation; the defect comparison unit of the insulator surface during the transmission line insulator defect detection is determined according to the located surface defect position and the layered surface defect state, and the transmission line insulator defect is detected according to the defect comparison unit. The above scheme is used to improve the comprehensiveness of the transmission line insulator defect detection under the influence of the deviation of the defect positioning and evaluation results.

[0036] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods. Figure 1 As shown, the figure is a module structure diagram of a transmission line insulator defect detection system based on multispectral imaging according to the present application and the present embodiment. The defect detection system includes: an image acquisition module 100, a defect feature determination module 200, a defect position positioning module 300, a defect state stratification module 400 and a defect comparison detection module 500, which are respectively described as follows:

[0037] The image acquisition module 100 is used to obtain a surface multispectral image of the transmission line insulator by setting the same data acquisition range for each multispectral imager and performing multispectral imaging data acquisition on the surface of the insulator to be detected on the transmission line.

[0038] In the specific implementation, it is necessary to first set the consistency of the multispectral imager to ensure that the spectral band range of all instruments is unified (such as 400-1000nm, covering the visible light to near-infrared range), and calibrate parameters such as exposure time, gain, and the number of spectral channels to eliminate differences between devices. Next, for the transmission line insulators, select an appropriate data acquisition method. When collecting on the ground, fix the imager on a high-stability platform, adjust the imaging angle to be perpendicular to the surface of the insulator, and ensure that the spectral reflection signal intensity is uniform; for areas that are difficult to access at high altitudes, deploy drones carrying multispectral imagers to collect data at close range according to the preset route, while maintaining flight stability and light source consistency to reduce the impact of external interference on imaging. After acquisition, the data is preprocessed, including spectral correction using the standard whiteboard reflectance curve to eliminate the deviation of the equipment and environmental conditions on the data, and Gaussian filtering to remove imaging noise and enhance image contrast, which will not be repeated here.

[0039] It should be noted that in this application, the surface multispectral image refers to a set of images formed by the reflection or emission spectrum information of the object surface at multiple specific wavelengths obtained by multispectral imaging technology. Each image corresponds to a wavelength range (spectral channel), which reflects the optical properties of a specific area of ​​the object surface at that wavelength.

[0040] The defect feature determination module 200 is used to perform mutual projection segmentation on the surface multi-spectral image to obtain multiple types of insulator defect feature information, and determine defect feature channels of different levels in the insulator defect image through all the insulator defect feature information.

[0041] In this embodiment, the surface multi-spectral image is segmented and the defect feature information of multiple types of insulators is obtained by the following steps:

[0042] Extract specific bands from multispectral images of transmission line insulator surfaces;

[0043] Determine the region of interest when interactively mapping the multispectral image according to the specific band;

[0044] The surface multispectral image is segmented into multiple types of insulator defect feature information through the region of interest.

[0045] In the specific implementation, first, all bands of the surface multispectral image are statistically analyzed, and the spectral difference between the normal area and the defective area of ​​each band on the insulator surface (such as the significance of the reflectivity or absorption peak) is calculated. Then, the spectral feature extraction method (such as principal component analysis PCA) is used to identify the specific band that can distinguish the defect category to the greatest extent. For example, if the reflectivity of the aging defect area decreases significantly in the 700nm band, this band can be regarded as a specific band. Then, the effectiveness of the specific band is verified, and the extracted band is evaluated through experiments to see whether it can stably distinguish different types of defects; then, the specific band is used to generate the initial region of interest (ROI, Region of Interest). The specific methods include: threshold segmentation, in the specific band image, a reflectivity or light intensity threshold is set to extract possible defect areas; edge detection: using Canny or Sobel operators to detect shape edges in the image and preliminarily locate the defect area; combining the spectral features of multiple bands, the region of interest is further refined through interactive mapping (i.e., information fusion between multiple bands). For example, the spectral similarity or difference between specific bands can be used to optimize the ROI range; then non-target areas (such as background, reflected light spots) can be removed to improve the pertinence and accuracy of the area; finally, clustering-based segmentation can be performed: K-means or Mean Shift algorithm can be applied to classify the pixels in the region of interest according to the spectral characteristics. Each type of pixel represents a defect feature; deep learning-based segmentation: using a deep learning model (such as U-Net), the region of interest is used as input and segmented into fine-grained defect categories; morphological optimization: post-processing the segmentation results, repairing noise or small errors (such as opening and closing operations to clean up artifacts), and then mapping the segmentation results back to the original image, annotating various defect feature information, such as aging areas, crack areas, contaminated areas, etc.

[0046] It should be noted that in the present application, the specific band refers to the specific wavelength range in the multispectral image that plays a key role in distinguishing normal areas from defective areas; the region of interest refers to the key analysis area selected according to specific targets or features in image analysis, which is used to concentrate resources for detailed processing; the defect feature information refers to the specific information set reflecting the defect type and nature in the image area extracted by segmentation, such as the size, shape and spectral characteristics of the crack area.

[0047] Preferably, in this embodiment, defect feature channels at different levels in the defect image of the insulator are determined by using all defect feature information of the insulator, referring to Figure 2 As described above, this figure is a schematic diagram of the process of determining a defect feature channel in some embodiments of the present application. In this embodiment, determining the defect feature channel can be implemented by the following steps:

[0048] In step S21, the characteristic coordinates of key points on the transmission line insulators are determined according to the characteristic information of all insulator defects;

[0049] In step S22, the position offset of the surface defect on the transmission line insulator is determined by the characteristic coordinates;

[0050] In step S23, defect identification information of the surface defect of the insulator is generated according to the position offset;

[0051] In step S24, defect feature channels at different levels in the defect image of the insulator are determined according to the defect identification information.

[0052] In the specific implementation, first, the center position, edge contour and geometric shape of the defect feature are extracted from the segmentation result of the multispectral image, and then the representative points of each type of defect are selected as key points by combining the spectral features and spatial features, such as the end point of the crack, the center point of the contaminated area, etc., and then the actual position of the insulator in the three-dimensional space is mapped to the characteristic coordinates in the two-dimensional image coordinate system by using the geometric transformation method (such as projection transformation); then, based on the comparison between the key point position and the standard geometric model of the insulator, the offset is measured, and the displacement of the defect area is estimated in the three-dimensional space using the depth information or stereo vision, and for each defect key point, its distance to the center axis of the normal surface or the reference point is calculated to form an offset vector, and then the offset is associated with the spectral feature, and the defect impact range and severity are evaluated by combining the band features; then, the defect feature coordinates, offset and multispectral features are combined to generate The defect description contains multi-dimensional information, for example: defect location: the specific coordinates or area on the surface of the insulator, defect type: classification based on multi-spectral features, such as aging, pollution, cracks, etc., defect severity: evaluation index generated based on offset, area and spectral characteristics, and the defect identification information is organized into a standardized data format for subsequent analysis and storage; finally, based on the defect identification information, the defect characteristics are hierarchically divided: the defect status is graded using hierarchical rules (such as slight, moderate, and severe), and each level corresponds to a defect category feature channel; feature channels are constructed: the spectral features of defects at each level are extracted to generate corresponding feature channels, for example: severe defect channels contain high reflectivity change areas, and slight defect channels contain low reflectivity change areas; the feature channel combination is optimized, and the multi-channel feature correlation is learned using a model (such as a convolutional neural network CNN).

[0053] It should be noted that in the present application, the characteristic coordinates are selected by analyzing the defect area to reflect the defect location and geometric characteristics of the coordinate points, usually center points, edge points or specific morphological points; the position offset is the spatial offset between the defect key point and the standard surface position, reflecting the displacement or deformation degree of the defect; the defect identification information represents a comprehensive description of the characteristics such as the location, type and severity of the insulator defect, which serves as the core data for subsequent processing; the defect feature channel refers to the channel extracted in the multispectral image according to the hierarchical division of the defect features, which is used to represent defect information of different severity or categories.

[0054] The defect position locating module 300 is used to obtain defect association information on the surface of the transmission line insulator, determine the defect assessment index of the defect position of the transmission line insulator according to the defect association information, and then locate the surface defect position of the transmission line insulator according to the defect assessment index.

[0055] In specific implementation, the defect association information on the surface of the transmission line insulator can be obtained in the following way, namely: first, extract the defect feature information from the multispectral image, including multidimensional data such as the position, shape, spectral characteristics and offset of each defect area. Secondly, establish the spatial and spectral association between defects, calculate the adjacent relationship of the defect area in the geometric position through the neighborhood analysis algorithm (such as Voronoi diagram or Delaunay triangulation), and evaluate the spectral correlation of the defect area in combination with the spectral similarity (such as cosine similarity or Euclidean distance). Then, use machine learning methods (such as neural network GNN in the figure) to further integrate this information, generate a defect association map, and mark the influence range of each defect and the coupling relationship with other defects. For example, identify the interaction between the contaminated area and the crack boundary or the gradual relationship between the aging area and other areas on the insulator surface. Finally, the results are standardized and output as a defect association information data set, which will not be repeated here.

[0056] It should be noted that in this application, defect correlation information refers to comprehensive information obtained by analyzing the spatial relationship, spectral characteristic correlation and mutual influence between different defect areas on the surface of transmission line insulators. It includes the positional proximity, spectral similarity, distribution pattern and coupling characteristics between defects, which is used to describe the possible interactions and potential causal relationships between defect areas.

[0057] Preferably, in this embodiment, the defect assessment index of the defect position of the transmission line insulator is determined according to the defect association information, referring to Figure 3 The figure is a schematic diagram of a process for determining a defect assessment index in some embodiments of the present application. In this embodiment, the defect assessment index can be determined by the following steps:

[0058] In step S31, the number of defect characteristic channels of the defect position of the transmission line insulator is determined according to the defect association information;

[0059] In step S32, feature fusion is performed on the number of defect feature channels to obtain adjustable parameters of the defect feature;

[0060] In step S33, extracting target contour information of the defect position of the transmission line insulator;

[0061] In step S34, a defect assessment index of a defect position of an insulator of a transmission line is determined according to the adjustable parameter and the target profile information.

[0062] In the specific implementation, first, according to the defect association information, all the characteristic channels involved in the defect area are counted, including spectral characteristic channels, geometric characteristic channels (such as offsets) and associated characteristic channels (such as coupling strength between adjacent defects), and the defect area is feature aggregated. By analyzing the distribution of the characteristic channels to which it belongs, the number of channels involved in describing the defect is calculated. For example, if a crack area shows spectral anomalies (spectral channels) and obvious geometric offsets (geometric channels) in a specific band, the number of its defect characteristic channels is 2, and the number of channels is normalized to facilitate subsequent feature fusion; then a multi-dimensional feature fusion algorithm (such as weighted fusion or attention mechanism) can be used to integrate the information of each channel: based on the weight of the associated information, different importance weights are assigned to the spectral, geometric, and associated features; deep learning methods (such as multi-layer perceptron MLP) are used to optimize the fusion strategy, generate comprehensive features, and extract adjustable parameters of the defect through the fusion results, such as the influencing factors of the crack depth or the influencing weights of the severity of contamination, which are used to quantify the overall impact of the defect. force; then, the contour information of the defect area is obtained from the defect segmentation result, including edge coordinates, shape features (such as perimeter, area) and texture distribution, and the contour is corrected using morphological methods (such as expansion and corrosion), noise is eliminated and the continuity of the contour is ensured, and the contour information is extracted, such as the length and width of the crack, the area ratio of the contaminated area, etc., as the evaluation input; finally, according to the adjustable parameters and the target contour information, a defect assessment model is constructed: the weight of the adjustable parameters is combined with the features in the target contour information (such as area, shape complexity), a comprehensive score is generated, and the evaluation formula is applied: for example, defect assessment index = (parameter weight × spectral channel characteristics) + (contour area × geometric factor). The calculation results are normalized so that the defect assessment index is within the set range (such as 0 to 1), which is convenient for comparing the severity of defects of different insulators, and then the evaluation results are output, and the corresponding severity classification (such as mild, moderate, severe) is marked in combination with the defect location.

[0063] It should be noted that in this application, the number of defect feature channels is the total number of multidimensional channels that characterize the defect area, reflecting its complexity in spectral, geometric or correlation characteristics; the adjustable parameter represents the quantitative index generated by feature fusion, which is used to adjust and optimize the defect evaluation results and reflect the comprehensive impact of the defect. The target contour information refers to the edge and geometric feature information of the defect area, which is used to describe the spatial morphology and distribution characteristics of the defect; the defect assessment index is a comprehensive quantitative index calculated based on the multidimensional characteristics and morphological information of the defect, which is used to reflect the severity and influence of the defect.

[0064] In specific implementation, the surface defect position of the transmission line insulator can be located by the defect assessment index in the following manner, namely: first, the defect assessment index is mapped to the coordinate system of the insulator surface, and each defect assessment point is geometrically corrected using the insulator structure model to ensure that the positioning point is consistent with the actual defect area. Next, the defect severity is classified (such as mild, moderate, and severe) according to the numerical range of the defect assessment index, and each category is assigned an independent color code or symbol identifier to visualize the defect position distribution. Secondly, multispectral imaging and three-dimensional reconstruction technology are combined to further accurately locate the three-dimensional spatial position of the defect, especially the irregularly distributed defect area on the surface of the complex insulator. Finally, a positioning optimization algorithm (such as the centroid positioning method or the regional growing algorithm) is applied to ensure the consistency of each defect position between the multispectral and three-dimensional models, and output standardized defect position coordinate data.

[0065] It should be noted that in this application, the surface defect position refers to the specific spatial position of the defect (such as cracks, aging, pollution, etc.) on the surface of the transmission line insulator relative to the geometric center of the insulator or the reference coordinate system. The position can be accurately represented by coordinates (such as two-dimensional image coordinates or three-dimensional space coordinates), and is usually described in combination with characteristic information such as the boundary, shape, depth, etc. of the defect.

[0066] The defect state stratification module 400 is used to set the anchor frame size of multispectral imaging during transmission line insulator defect detection, determine the defect prediction deviation of the transmission line insulator defect position according to the anchor frame size and the defect feature channel, and stratify the surface defect state of the transmission line insulator according to the defect prediction deviation.

[0067] In specific implementation, the anchor frame size of multispectral imaging for transmission line insulator defect detection can be set in the following way, namely: first, the anchor frame size needs to be adjusted according to the actual size and defect characteristics of the insulator. Considering the size and shape differences of defects on the surface of the insulator, the scale range of the possible defect area is determined through prior knowledge or through the defect distribution in the training data set. Different anchor frame sizes are used to cover these defect areas of different scales to ensure that each type of defect can be accurately captured. For example, for cracks or small cracks, the anchor frame should be set to a smaller size, while for aging or large areas of pollution, the anchor frame should be set to a larger size; then, through experimental data or simulation models, combined with the reflection characteristics of different bands in the multispectral image, the size of each anchor frame is optimized. Using deep learning methods such as region proposal network (RPN), candidate anchor frames with different sizes can be automatically generated in the image, which can cover potential defect areas. The model is trained to fine-tune the selection of anchor frame size so that each defect can be captured to the greatest extent within its corresponding anchor frame. Finally, the matching degree between the anchor frame size and the defect feature is evaluated, and the size configuration of the anchor frame is further optimized in combination with its response intensity in multispectral images.

[0068] It should be noted that in this application, the anchor box size refers to the size and shape of the candidate box in the target detection algorithm to identify and locate the target object (such as defects, target areas, etc.) in the image. The anchor box size is usually set based on the object size distribution in the input image and automatically adjusted by the algorithm to accommodate objects of different scales. In multispectral imaging and defect detection, the anchor box size determines the size of the frame used to cover the defect area, usually expressed in pixel values ​​of width and height.

[0069] In this embodiment, the defect prediction deviation of determining the defect position of the transmission line insulator according to the anchor frame size and the defect characteristic channel can be implemented by the following steps:

[0070] Determining defect test information of the transmission line insulator by means of the anchor frame size;

[0071] Determine the overlap sensitivity of the defect position of the transmission line insulator according to the defect characteristic channel;

[0072] The defect prediction deviation of the defect position of the transmission line insulator is determined according to the defect test information and the overlap sensitivity.

[0073] In the specific implementation, first, the image is scanned according to the set anchor frame size (for example, width and height) to locate each potential defect area; for each anchor frame, the pixel data it contains is evaluated to determine whether these pixels belong to the defect area (such as cracks, stains, etc.). This step requires the use of image segmentation or classification algorithms to calibrate whether each anchor frame is valid, and then calculate the spectral characteristics, shape characteristics, etc. of the defect area corresponding to each anchor frame for subsequent deviation calculation. By comparing the error between the actual defect position and the estimated position of the anchor frame, defect test information is formed, including the size, position and morphological characteristics of the defect; then, the overlap can be quantified by calculating the "intersection over union" (IoU). The higher the IoU value, the greater the overlap between the anchor frame and the actual defect area, and the higher the sensitivity. The overlap between each anchor frame and the actual defect area is calculated, and the credibility of the defect prediction is adjusted according to the degree of overlap. For example, an area with a high degree of overlap can be considered to have a more accurate prediction of the anchor frame for that area, while an area with a low degree of overlap may need further optimization. Perform an overall analysis of multiple anchor frames and calculate the average overlap sensitivity, which will directly affect the adjustment of subsequent defect prediction deviations; finally, use the overlap between the anchor frame and the actual defect, combined with defect test information (such as the defect's feature channel), to calculate the prediction deviation. This deviation reflects the gap between the estimated position of the anchor frame and the actual defect position. According to the size of the prediction deviation, adjust the anchor frame position to make it closer to the actual defect position. The specific adjustment process can be achieved through optimization algorithms (such as gradient descent), the goal of which is to minimize the prediction deviation, correct the prediction deviation through multiple iterations, and finally determine the most accurate defect position.

[0074] It should be noted that in this application, defect test information refers to the data set about defect features obtained by analyzing candidate areas (such as anchor boxes) during defect detection; overlap sensitivity refers to the degree of overlap between anchor boxes and actual defect areas during defect prediction; defect prediction deviation refers to the error between the defect position, size or features predicted by the algorithm and the actual position, size or features of the actual defect during defect detection. This deviation reflects the accuracy and reliability of the detection model in identifying and locating defects.

[0075] In specific implementation, the stratification of the surface defect states of the transmission line insulators according to the defect prediction deviation can be achieved in the following manner, namely: first, the defect prediction deviation is compared with the predefined threshold range, and the deviation value is corresponded to the specific defect level. For example, the area with a smaller prediction deviation can be marked as "minor defect", the medium deviation as "medium defect", and the larger deviation as "serious defect", thereby forming a preliminary stratification basis. Next, the stratification results are further optimized by analyzing the spatial distribution and spectral characteristics of the defects in each layer. For example, the position deviation information in the prediction deviation is used to accurately locate the defect area, and the size deviation is combined to determine whether the defect range is consistent with the expectation. For areas with large feature deviations, further analyze whether the spectral characteristics change significantly to ensure accurate identification of serious defects. Finally, a machine learning classifier or a rule-based grading algorithm is used to mark the status of all defect areas to form the final stratification result.

[0076] It should be noted that in this application, the surface defect state refers to the classification results of the current physical characteristics and deterioration degree of the surface defects of the transmission line insulators, reflecting the potential impact of the defects on the insulator function. This state is usually described in layers based on the type of defect (such as cracks, contamination, aging), severity (mild, moderate, severe), spatial distribution (local or comprehensive) and morphological characteristics (depth, area, spectral characteristics, etc.). The classification of surface defect states aims to quantify the impact of surface damage on insulators.

[0077] The defect comparison detection module 500 is used to determine the defect comparison unit of the insulator surface during the transmission line insulator defect detection according to the located surface defect position and the surface defect state after stratification, and then detect the transmission line insulator defects according to the defect comparison unit.

[0078] In specific implementation, the defect comparison unit of the insulator surface during the detection of transmission line insulator defects is determined according to the surface defect position after positioning and the surface defect state after layering, and then the detection of transmission line insulator defects according to the defect comparison unit can be implemented in the following manner, namely: first, based on the surface defect position after positioning, the insulator surface is divided into a number of comparison units, which can be regular grids or adaptive regions (dynamically adjusted according to the distribution characteristics of the defects). Each comparison unit is associated with the defect coordinates and morphological boundaries after positioning to ensure that the unit can accurately cover the defect area. For example, the narrow and long area where the crack is located can be set as an irregular polygonal unit, and the contaminated area can be set as a rectangular or elliptical unit. Next, combined with the surface defect state after layering, the defect state weight is introduced in each comparison unit. For example, the weight of the slight defect state is low, while the weight of the severe defect state is high. By embedding the layering information into the comparison unit, the defect detection strategy can be further refined. Then, the spectral features, geometric features and texture features are extracted in the comparison unit, compared with the features in the standard defect template library, and the similarity score is calculated to evaluate the type and severity of the defect. This step can be achieved by using algorithms based on feature distance (such as Euclidean distance) or deep learning methods (such as feature vector matching). Finally, the detection results of the comparison units are summarized to build an overall defect detection report, and the detected defect results are mapped to the insulator surface model to generate a visual defect distribution map.

[0079] It should be noted that in this application, the defect comparison unit refers to a specific area divided for accurate comparison and evaluation during the defect detection of transmission line insulators. It is based on a certain position or area on the surface of the insulator, combined with the spatial positioning information, characteristic attributes (such as shape, spectral characteristics, texture, etc.) and layered state of the defect, and compares and analyzes the defect characteristics in the area with the standard reference template.

[0080] It can be seen that in the present application, it is possible to detect defects in insulators of transmission lines under the influence of deviations in defect positioning and evaluation results; wherein, by setting the same data acquisition range and performing multispectral image acquisition under the same standard, it is possible to ensure that consistent and high-quality image data is obtained, which helps to reduce the deviation caused by inconsistent imager configuration and improve the stability and reliability of subsequent defect detection results; by accurately extracting multiple types of defect feature information on the surface of the insulator through the mutual reflection segmentation technology, it is possible to more comprehensively identify different types of defects, thereby improving the defect classification and recognition accuracy of the system, and avoiding the problems of insufficient image segmentation accuracy and inaccurate defect extraction in the prior art; by acquiring and analyzing defect association information, it is possible to classify and identify defects based on the actual location and characteristics of the defects. The feature generates an accurate defect assessment index, making defect assessment more scientific and reliable, thereby improving the accuracy of defect positioning and overcoming the technical defect of low assessment accuracy in the prior art; setting a reasonable anchor frame size and optimizing the prediction deviation in combination with the defect feature channel can effectively improve the accuracy of defect detection, especially when the defect size and position are difficult to determine, thereby improving the problem of large deviation in the prior art and improving the accuracy of defect prediction; by setting a defect comparison unit based on the surface defect position after positioning and the defect state after stratification, the surface defects of the insulator can be accurately compared and detected one by one, improving the hierarchical accuracy and automation level of defect detection, and further overcoming the problem of lack of accurate comparison and intelligent stratification in the existing detection system.

[0081] In summary, the technical solution adopted in the present application can improve the comprehensiveness of transmission line insulator defect detection under the influence of deviations in defect location and evaluation results.

[0082] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0083] Those skilled in the art will appreciate that all or part of the steps in the various methods of the above embodiments may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, the storage medium including a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically-erasable programmable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0084] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

Claims

1. A transmission line insulator defect detection system based on multispectral imaging, characterized in that: The defect detection system comprises: An image acquisition module is used to obtain a surface multispectral image of the transmission line insulator by setting the same data acquisition range for each multispectral imager and performing multispectral imaging data acquisition on the surface of the insulator to be detected on the transmission line; A defect feature determination module is used to perform mutual projection segmentation on the surface multi-spectral image to obtain multiple types of insulator defect feature information, and determine defect feature channels of different levels in the defect image of the insulator through all the insulator defect feature information; A defect location positioning module, used to obtain defect association information on the surface of the transmission line insulator, determine a defect assessment index of the defect position of the transmission line insulator according to the defect association information, and then locate the surface defect position of the transmission line insulator according to the defect assessment index; A defect state stratification module is used to set the anchor frame size of multispectral imaging during transmission line insulator defect detection, determine the defect prediction deviation of the transmission line insulator defect position according to the anchor frame size and the defect feature channel, and stratify the surface defect state of the transmission line insulator according to the defect prediction deviation; The defect comparison detection module is used to determine the defect comparison unit on the insulator surface during transmission line insulator defect detection according to the surface defect position after positioning and the surface defect state after stratification, and then detect the transmission line insulator defects based on the defect comparison unit.

2. A transmission line insulator defect detection system based on multispectral imaging as claimed in claim 1, characterized in that: The surface multispectral image refers to an image set formed by the reflection or emission spectrum information of the object surface at multiple specific wavelengths obtained by multispectral imaging technology.

3. A transmission line insulator defect detection system based on multispectral imaging as claimed in claim 1, characterized in that: The surface multi-spectral image is segmented and mapped to obtain multiple types of insulator defect feature information, including: Extract specific bands from multispectral images of transmission line insulator surfaces; Determine the region of interest when interactively mapping the multispectral image according to the specific band; The surface multispectral image is segmented into multiple types of insulator defect feature information through the region of interest.

4. A transmission line insulator defect detection system based on multispectral imaging as claimed in claim 1, characterized in that: The defect feature information refers to a specific information set reflecting the defect type and nature in the image region extracted by segmentation.

5. The transmission line insulator defect detection system based on multispectral imaging according to claim 1, characterized in that: The defect feature channel refers to a channel extracted according to the hierarchical division of defect features in a multispectral image.

6. A transmission line insulator defect detection system based on multispectral imaging as claimed in claim 1, characterized in that: Defect correlation information refers to the comprehensive information obtained by analyzing the spatial relationship, spectral characteristic correlation and mutual influence between different defect areas on the surface of transmission line insulators.

7. A transmission line insulator defect detection system based on multispectral imaging as claimed in claim 1, characterized in that: The surface defect position refers to the specific spatial position of the defect on the surface of the transmission line insulator relative to the geometric center of the insulator or the reference coordinate system.

8. The transmission line insulator defect detection system based on multispectral imaging according to claim 1, characterized in that: The defect prediction deviation of determining the defect position of the transmission line insulator according to the anchor frame size and the defect characteristic channel specifically includes: Determining defect test information of the transmission line insulator by means of the anchor frame size; Determine the overlap sensitivity of the defect position of the transmission line insulator according to the defect characteristic channel; The defect prediction deviation of the defect position of the transmission line insulator is determined according to the defect test information and the overlap sensitivity.

9. The transmission line insulator defect detection system based on multispectral imaging according to claim 1, characterized in that: The surface defect state refers to the classification result of the current physical characteristics and deterioration degree of the surface defects of the transmission line insulator.

10. The transmission line insulator defect detection system based on multispectral imaging according to claim 1, characterized in that: The defect comparison unit refers to a specific area divided for achieving accurate comparison and evaluation during the defect detection process of insulators of transmission lines.

Images