A machine vision-based power fitting surface intelligent monitoring method and system

Abstract

The application relates to the technical field of electric power fitting monitoring, and discloses an intelligent electric power fitting surface monitoring method and system based on machine vision, which comprises the following steps: acquiring an electric power fitting image; analyzing the electric power fitting image to obtain electric power fitting surface color data of the electric power fitting image; acquiring a shooting position of an inspection image, and determining material parameters of the electric power fitting according to the shooting position; determining the rust color of the electric power fitting according to the material parameters of the electric power fitting, comparing the color parameters with the rust color, determining whether rust occurs, and if rust occurs, determining the real-time rust degree of the electric power fitting according to the color area; determining the predicted rust degree according to meteorological data and the material parameters of the electric power fitting, and judging the rust severity grade of the electric power fitting surface according to the real-time rust degree and the predicted rust degree. The application can realize accurate detection and quantitative evaluation of the rust degree of the electric power fitting surface.

Description

Technical Field

[0001] This invention relates to the field of power fitting monitoring technology, and more specifically, to a machine vision-based intelligent monitoring method and system for the surface of power fittings. Background Technology

[0002] Power fittings, as critical connecting components in power transmission lines, are widely used in scenarios such as conductor fixing, connection, and protection. Their operational status directly affects the safety and stability of power transmission lines. During long-term outdoor service, power fittings are susceptible to corrosion due to complex meteorological environmental factors such as temperature and humidity, rainfall, salt spray, and industrial exhaust gases. Corrosion leads to a decrease in the mechanical strength and conductivity of the fittings, and in severe cases, may cause safety accidents such as fitting breakage and short circuits, resulting in huge economic losses and social impact.

[0003] Currently, corrosion monitoring of power fittings mainly relies on manual inspection. Inspectors visually inspect the surface of the fittings or use simple tools, but this method has several drawbacks: First, it is inefficient, as transmission lines span large areas and are widely distributed, requiring significant manpower, resources, and time for manual inspection. Second, it has low accuracy, as subjective factors such as the inspector's experience, sense of responsibility, and visual fatigue make it difficult to accurately determine the degree and trend of corrosion, easily leading to missed or false detections. Third, it poses safety risks, as some transmission lines are located in complex terrain such as mountains and valleys, posing risks of falls and electric shock during manual inspection.

[0004] With the development of machine vision and artificial intelligence technologies, image-based automated monitoring methods are gradually being applied to the field of power equipment inspection. However, existing technologies still have shortcomings in monitoring corrosion of power fittings: on the one hand, there is a lack of precise target detection mechanisms for power fittings, making it difficult to efficiently separate the fitting areas from complex inspection images; on the other hand, the characteristics of fitting materials and environmental factors are not fully considered, and corrosion status is judged solely through single image color analysis, resulting in insufficient accuracy and reliability of corrosion detection, inability to effectively predict corrosion development trends, and difficulty in meeting the power industry's demand for precise monitoring of fitting conditions.

[0005] Therefore, there is an urgent need for an intelligent monitoring method that can accurately detect surface corrosion of power fittings, quantitatively assess the degree of corrosion, and predict development trends, in order to solve the problems of low efficiency, poor accuracy, and insufficient safety of existing monitoring methods and ensure the safe and stable operation of transmission lines. Summary of the Invention

[0006] In view of this, the present invention proposes a machine vision-based intelligent monitoring method and system for the surface of power fittings, aiming to solve the problems of low accuracy, poor precision and insufficient safety of traditional monitoring, which does not consider the material characteristics and environmental factors of the fittings and only relies on a single factor for judgment.

[0007] On the one hand, the present invention proposes a machine vision-based intelligent monitoring method for the surface of power fittings, including: acquiring inspection images of transmission lines, performing target detection on the inspection images, determining whether the inspection images contain power fittings, and if they contain power fittings, cropping the inspection images to obtain power fitting images;

[0008] The power fitting image is preprocessed and color analyzed to obtain the power fitting surface color data, which includes color parameters and color area.

[0009] The system acquires the shooting location of the inspection image, determines the installation type of the power fitting based on the preset power fitting database, and determines the material parameters of the power fitting based on the installation type.

[0010] The corrosion color of the power fitting is determined based on the material parameters of the power fitting. The color parameters in the surface color data of the power fitting are compared with the corrosion color to determine whether the power fitting has rusted. If corrosion is determined, the real-time corrosion degree of the power fitting is determined based on the color area in the surface color data of the power fitting.

[0011] Meteorological data of the corresponding location of the power fittings are obtained based on the shooting location. The corrosion reaction of the power fittings is predicted based on the meteorological data and the material parameters of the power fittings. The predicted corrosion degree of the power fittings is obtained. The severity level of corrosion on the surface of the power fittings is determined based on the real-time corrosion degree and the predicted corrosion degree.

[0012] Preferably, target detection is performed on the inspection image to determine whether the inspection image contains power fittings, specifically as follows:

[0013] The inspection images are input into a pre-trained power fittings target detection model;

[0014] Feature extraction is performed on the inspection image. Deep semantic features of the inspection image are obtained through convolutional layers and pooling layers. Candidate power fitting regions are generated using a region generation network. The candidate power fitting regions are classified and bounding box regressed to obtain the probability value of each candidate power fitting region belonging to power fitting and the corresponding precise bounding box coordinates.

[0015] When the probability value of a candidate power fitting area is greater than the preset confidence threshold, it is determined that the inspection image contains power fittings, and the coordinate information of all bounding boxes that meet the conditions is recorded.

[0016] Preferably, the image of the power fitting is preprocessed and color analyzed to obtain the surface color data of the power fitting, specifically as follows:

[0017] Illumination correction is performed on images of power fittings using the Retinex algorithm or white balance correction.

[0018] Image segmentation technology is applied to separate the power fittings from the background to obtain the image of the power fitting itself;

[0019] Color features are extracted from the image of the power fitting body. The image of the power fitting body is converted from RGB space to CIELAB color space. Dominant color clusters are extracted by K-means clustering. The average L*, a*, b* values ​​of each dominant color cluster and the proportion of the pixel area occupied are obtained to obtain color parameters and color area.

[0020] Preferably, based on a preset power fitting database, the installation type of the power fitting is determined according to the shooting location, and the material parameters of the power fitting are determined based on the installation type, specifically as follows:

[0021] The preset power fittings database includes information on several road segments, each segment corresponding to the installation type, quantity, and installation time of the power fittings.

[0022] Based on the shooting location, the corresponding road segment information is determined, and the preset power fitting database is filtered to determine the power fitting installation type corresponding to the shooting location;

[0023] The material parameters of the power fittings are determined according to the installation type of the power fittings. The material parameters include the metal composition of the fitting base, the type of surface coating, the color of the surface coating, and the thickness of the surface coating.

[0024] Preferably, the rust color of the power fitting is determined based on the material parameters of the power fitting, specifically as follows:

[0025] A material-rust color mapping table is preset, which includes several metal materials, and each metal material is given a standard rust color range under standard atmospheric corrosion conditions.

[0026] The metal composition of the hardware substrate is matched with the material-rust color mapping table to determine the standard rust color range corresponding to the match between the metal composition of the hardware substrate and the metal material.

[0027] Preferably, the color parameters in the surface color data of the power fittings are compared with the rust color to determine whether the power fittings have rusted. Specifically:

[0028] The color parameters are matched with the values ​​of the surface coating color. Color parameters that successfully match the values ​​of the surface coating color are discarded to obtain suspected rust color parameters.

[0029] The minimum color difference between the suspected rust color parameter and the standard rust color range is determined based on the suspected rust color parameter and the standard rust color range.

[0030] The minimum color difference is compared with the preset color difference. If the minimum color difference is less than or equal to the preset color difference, it is determined that the power fitting has rusted; if the minimum color difference is greater than the preset color difference, it is determined that the power fitting has not rusted.

[0031] Preferably, the real-time corrosion degree of the power fitting is determined based on the color area in the surface color data of the power fitting, specifically by setting the color area corresponding to each suspected corrosion color parameter as the suspected corrosion color area.

[0032] Calculate the sum of the suspected rust color areas, determine the difference between the sum of the suspected rust color areas and the total surface area of ​​the power fittings, and set the difference as the real-time rust degree of the power fittings.

[0033] Preferably, meteorological data of the corresponding location of the power fitting is obtained based on the shooting location, and the corrosion reaction of the power fitting is predicted based on the meteorological data and the material parameters of the power fitting to obtain the predicted corrosion degree of the power fitting, specifically as follows:

[0034] The installation time of the power fittings is determined based on a pre-set power fittings database. Meteorological data corresponding to the installation time and location of the power fittings are then obtained based on the installation time and shooting location. This meteorological data includes rainfall, temperature and humidity, precipitation frequency, and salt spray concentration. concentration;

[0035] Based on a pre-built corrosion prediction model, the predicted corrosion degree of power fittings is determined according to the material parameters and meteorological data.

[0036] Preferably, the severity of corrosion on the surface of the power fittings is determined based on the real-time corrosion degree and the predicted corrosion degree, specifically as follows:

[0037] The real-time corrosion degree is compared with the predicted corrosion degree. If the real-time corrosion degree is less than or equal to the predicted corrosion degree, it is determined that the corrosion on the surface of the power fittings is not serious.

[0038] If the real-time corrosion degree is greater than the predicted corrosion degree, then the corrosion degree difference between the real-time corrosion degree and the predicted corrosion degree is determined, and the corrosion severity level of the power fitting surface is judged based on the corrosion degree difference.

[0039] A first rust degree difference and a second rust degree difference are preset, wherein the first rust degree difference is less than the second rust degree difference;

[0040] The corrosion severity level of the power fitting surface is set according to the relationship between the corrosion degree difference and the first corrosion degree difference and the second corrosion degree difference;

[0041] If the corrosion degree difference is less than the first corrosion degree difference, then the corrosion severity level of the power fitting surface is set to Level 1 severe level.

[0042] If the corrosion degree difference is greater than or equal to the first corrosion degree difference, and the corrosion degree difference is less than the second corrosion degree difference, then the corrosion severity level of the power fitting surface is set to Level 2 severe level.

[0043] If the corrosion degree difference is greater than or equal to the second corrosion degree difference, the corrosion severity level of the power fitting surface is set to Level 3, wherein the severity of Level 1, Level 2, and Level 3 increases sequentially.

[0044] Compared with existing technologies, the advantages of this invention are as follows: This invention employs a pre-trained target detection model, which can accurately identify power fittings in inspection images and locate their bounding box coordinates, significantly improving the accuracy of target detection. By preprocessing and color analysis of the power fitting images, combined with the K-means clustering algorithm to extract the dominant color clusters and their average values, the color parameters and area of ​​the power fitting surface can be accurately obtained. By comparing with a preset standard corrosion color range, the presence and degree of corrosion of the power fittings can be accurately determined. Based on the installation location and time of the power fittings, corresponding meteorological data is obtained, and combined with material parameters using a pre-constructed corrosion prediction model, the corrosion rate trend of the power fittings can be effectively predicted, providing a scientific basis for preventative maintenance. The entire monitoring process is highly automated, from image acquisition and processing to the final assessment of corrosion severity, greatly reducing the need for manual intervention, improving work efficiency, and also reducing errors caused by human factors.

[0045] On the other hand, this application also provides a machine vision-based intelligent surface monitoring system for power fittings, used to apply the above-mentioned machine vision-based intelligent surface monitoring method for power fittings, including:

[0046] The power fittings image acquisition module is configured to acquire inspection images of transmission lines, perform target detection on the inspection images, determine whether the inspection images contain power fittings, and if they contain power fittings, crop the inspection images to obtain power fittings images.

[0047] The power fitting color analysis module is configured to preprocess and perform color analysis on the power fitting image to obtain power fitting surface color data, which includes color parameters and color area.

[0048] The power fitting material determination module is configured to acquire the shooting location of the inspection image, determine the installation type of the power fitting based on the shooting location according to the preset power fitting database, and determine the material parameters of the power fitting based on the installation type.

[0049] The power fitting corrosion determination module is configured to determine the corrosion color of the power fitting based on the material parameters of the power fitting, compare the color parameters in the surface color data of the power fitting with the corrosion color to determine whether the power fitting has rusted, and if corrosion is determined, determine the real-time corrosion degree of the power fitting based on the color area in the surface color data of the power fitting.

[0050] The corrosion severity judgment module is configured to acquire meteorological data of the corresponding location of the power fitting based on the shooting location, predict the corrosion reaction of the power fitting based on the meteorological data and the material parameters of the power fitting, obtain the predicted corrosion degree of the power fitting, and judge the corrosion severity level of the power fitting surface based on the real-time corrosion degree and the predicted corrosion degree.

[0051] It is understood that the defect detection method and system for drop-out fuse protective covers provided in this application have the same beneficial effects, and will not be described in detail here. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly described below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0053] Figure 1 A flowchart illustrating the intelligent surface monitoring method for power fittings based on machine vision provided in an embodiment of the present invention;

[0054] Figure 2 This is a functional block diagram of a machine vision-based intelligent surface monitoring method system for power fittings provided in an embodiment of the present invention. Detailed Implementation

[0055] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0056] In some embodiments of this application, see Figure 1 As shown, this embodiment provides a machine vision-based intelligent monitoring method for the surface of power fittings, including:

[0057] S100: Obtain an inspection image of the transmission line, perform target detection on the inspection image, determine whether the inspection image contains power fittings, and if it contains power fittings, crop the inspection image to obtain an image of power fittings.

[0058] Specifically, acquiring inspection images of power transmission lines involves using drones, robots, or other equipment to photograph the lines and collect image information, including power fittings. Target detection technology is applied to these images to identify and locate the power fittings. Once the inspection images confirm the presence of power fittings, the next step is to crop the portion containing the fittings to extract individual images for further detailed analysis and processing.

[0059] Specifically, to obtain a more comprehensive assessment of the condition of power fittings, multiple images can be taken of the same location from different angles (front, side, and top view), and then the data from these perspectives can be merged to provide more accurate detection and analysis. In addition to data from a single inspection, historical inspection records can be combined to construct a time-series dataset. By comparing image changes over different periods, long-term trends in power fittings can be monitored, providing early warnings of potential problems. Considering that different weather conditions may affect image quality, specific algorithms can be developed to optimize image processing under adverse weather conditions such as fog and rain, ensuring high-precision target detection capabilities in any environment.

[0060] S200, preprocess and perform color analysis on the image of the power fittings to obtain the surface color data of the power fittings in the image of the power fittings. The surface color data of the power fittings includes color parameters and color area.

[0061] Specifically, preprocessing and color analysis of power fitting images involves, after obtaining the cropped image of the power fitting, firstly eliminating interference factors such as uneven lighting and color cast through image enhancement techniques to restore the true color state of the fitting surface; then, using image segmentation technology to separate the fitting body from the background and extracting color features within its effective area. The final output of this process, "power fitting surface color data," includes two aspects: first, color parameters, which are numerical values ​​describing the essential attributes of color (such as L*, a*, and b* values ​​in the CIELAB color space), used to characterize the color type and hue; and second, color area, which is the proportion of pixels or actual physical area occupied by the color on the fitting surface, used to reflect the color coverage range.

[0062] Specifically, in addition to CIELAB, other color spaces such as HSV and YUV can be introduced for cross-validation to improve the robustness of color discrimination. Especially under complex lighting conditions or when coating aging causes color shifts, multi-space joint analysis helps improve the accuracy of rust identification. The calculation of color area can be combined with the geometric structure information of the hardware (such as the surface area corresponding to a known model) to convert pixel ratios into actual physical areas. Simultaneously, the color area threshold for judging rust can be dynamically adjusted based on different installation locations, orientations, or historical rust trends, achieving a more refined state assessment. Based on color analysis, surface texture features (such as roughness and patch continuity) are further integrated to help distinguish rusted areas from non-rusted discoloration areas such as stains and oil stains, avoiding misjudgments. The color data obtained this time is compared with the color parameters and areas in historical inspection records to construct a color change trajectory, used to identify the speed and pattern of rust development, providing richer input dimensions for the prediction model.

[0063] S300: Obtain the shooting location of the inspection image, determine the installation type of the power fitting based on the shooting location according to the preset power fitting database, and determine the material parameters of the power fitting based on the installation type.

[0064] Specifically, obtaining the shooting location of the inspection images refers to determining the specific geographical location of the transmission line inspection images using GPS or other positioning technologies. Based on a pre-set database of power fittings, which contains detailed information on power fittings installed in different sections and locations, such as installation type (e.g., suspension clamp or tension clamp), installation time, and material composition, the corresponding power fitting information can be filtered from the database based on the shooting location of the inspection images, and its installation type can be determined accordingly. Furthermore, based on the known installation type, the specific material parameters used in that type of power fitting can be queried or calculated, including the base metal composition, surface coating type, color, and thickness.

[0065] Specifically, in addition to static database queries, a dynamic update mechanism can be established to ensure that new data obtained after each inspection (such as newly discovered hardware models or material parameters) is fed back and updated to the database in real time. This not only improves the accuracy and completeness of the database but also provides more precise support for future inspections. Consideration should also be given to integrating data from other sources (such as meteorological and geological data) with the power hardware database. For example, material selection recommendations for power hardware could be adjusted based on local climate conditions, or special protective measures could be implemented for power facilities in certain areas based on geological activity predictions.

[0066] S400, determine the corrosion color of the power fitting based on the material parameters of the power fitting, compare the color parameters in the surface color data of the power fitting with the corrosion color to determine whether the power fitting has rusted, and if corrosion is determined, determine the real-time corrosion degree of the power fitting based on the color area in the surface color data of the power fitting.

[0067] Specifically, the system combines materials science knowledge with image color analysis to intelligently assess the corrosion status of power fittings. First, based on known material parameters of the power fittings (such as the base metal composition and surface coating type), the system retrieves the standard corrosion color range that the material might exhibit under typical atmospheric corrosion conditions from a pre-defined "material-corrosion color mapping table" (e.g., iron-based materials typically appear reddish-brown, while copper alloys may appear green or bluish-green). Then, the surface color parameters of the power fittings extracted in the previous steps (such as L*, a*, and b* values ​​in CIELAB space) are compared with this standard corrosion color range. The system determines whether a matching corrosion characteristic color exists by calculating color difference. If a match is found, corrosion is determined. Further, using the pixel area corresponding to the corrosion color (i.e., color area) combined with the overall surface area of ​​the fitting, a "real-time corrosion degree" is quantified to reflect the severity of the current corrosion coverage.

[0068] Specifically, to address the phenomenon that the same material may exhibit different colors at different corrosion stages (initial oxidation, intermediate loose rust layer, and late-stage peeling), the material-rust color mapping table can be refined into multiple color subsets corresponding to different corrosion stages. This allows for the determination of the corrosion development stage and enhances the assessment dimensionality. A coating aging and color change model is introduced to distinguish non-corrosive color changes caused by ultraviolet radiation, temperature variations, etc. (such as yellowing or fading of the coating) from the true rust color, avoiding misjudgment. For example, aging compensation rules can be established by comparing the color shift between the original coating color and the current non-rusted area. Considering the significant differences in the geometry of different hardware models, a normalization method based on a 3D model or standard projected area can be introduced to convert pixel area into physically meaningful rust coverage (such as mm² or percentage), improving the consistency and comparability of rust degree assessment. Based on the identified rusted areas, further analysis of the edge clarity, distribution concentration, or spread trend of rust patches (such as dot-like, patchy, or edge-priority patterns) serves as an auxiliary indicator for judging rust activity or potential failure risk, enriching the description of the rust state.

[0069] S500: Based on the shooting location, obtain meteorological data of the corresponding location of the power fitting; predict the corrosion reaction of the power fitting based on the meteorological data and the material parameters of the power fitting; obtain the predicted corrosion degree of the power fitting; and determine the severity level of corrosion on the surface of the power fitting based on the real-time corrosion degree and the predicted corrosion degree.

[0070] Specifically, by combining environmental factors with material properties, a predictive capability for the corrosion evolution of power fittings is constructed. First, based on the location of the inspection images (such as latitude, longitude, and altitude), the system obtains historical and recent meteorological data for that location during the service life of the power fittings from meteorological databases or environmental monitoring platforms. This data includes, but is not limited to, rainfall, relative humidity, temperature changes, precipitation frequency, salt spray concentration, and sulfur dioxide (SO2) concentration. The content of corrosive gases such as metals and coatings is then measured. Subsequently, known material parameters of power fittings (such as metal type, coating type and thickness) are input into a pre-constructed corrosion prediction model. This model, based on material corrosion kinetics principles or data-driven methods, simulates the electrochemical or chemical corrosion processes that may occur on the surface of the fittings under specific meteorological conditions, thus outputting a "predicted corrosion degree"—the theoretically expected degree of corrosion under the current environment and material combination. Finally, the actual detected "real-time corrosion degree" is compared with this predicted value. If the actual corrosion significantly exceeds expectations, it indicates abnormal accelerated corrosion or other potential risks. Based on this, the corrosion severity level is classified (e.g., Level 1, Level 2, Level 3), providing a graded early warning basis for operation and maintenance decisions.

[0071] Specifically, the time span of meteorological data can be dynamically selected based on the installation time of the fittings (e.g., the past year, the past three years, or the entire service life), and the impact of different time periods can be weighted (e.g., recent weather has a greater impact on the current state), improving the timeliness and accuracy of the prediction model. Based on the basic meteorological parameters, microclimate factors (e.g., local shading effect, differences in sunlight due to orientation, dust / water accumulation areas) or pollution source distribution (e.g., emissions from nearby chemical plants) can be further introduced to construct a more refined regional corrosion environment profile, enhancing the scenario adaptability of the prediction. The deviation between each "real-time corrosion degree" and "predicted corrosion degree" is recorded and used for online optimization or periodic updates to the corrosion prediction model parameters, enabling continuous learning and self-correction of the model and improving long-term prediction reliability. The threshold for judging the corrosion degree difference (e.g., the first and second differences) can be configured differently based on material type, service life, or line importance. For example, stricter thresholds are used for critical transmission channels or aging fittings to improve early warning sensitivity. In addition to outputting the current predicted corrosion level, it can also generate corrosion evolution curves for the next 6 months or 1 year, and display them overlaid with safety threshold lines to assist maintenance personnel in developing forward-looking maintenance plans.

[0072] In some embodiments of this application, target detection is performed on the inspection image to determine whether the inspection image contains power fittings. Specifically, the inspection image is input into a pre-trained power fitting target detection model; feature extraction is performed on the inspection image, obtaining deep semantic features of the inspection image through convolutional layers and pooling layers; candidate power fitting regions are generated using a region generation network; and the candidate power fitting regions are classified and bounding box regressed to obtain the probability value of each candidate power fitting region belonging to power fittings and the corresponding precise bounding box coordinates; when the probability value of a candidate power fitting region is greater than a preset confidence threshold, it is determined that the inspection image contains power fittings, and the bounding box coordinate information of all satisfying conditions is recorded. If no fittings are detected, the image is skipped and recorded as a "no fittings region".

[0073] Specifically, this solution can automatically and accurately identify the presence of power fittings in inspection images and precisely locate their position in the image, providing a reliable basis for subsequent image cropping and surface condition analysis.

[0074] Specifically, the acquired inspection images are input into a pre-trained power fitting target detection model. This model is built on a deep learning architecture and has the ability to identify specific targets from images. Internally, the model processes the input image step-by-step through multiple convolutional and pooling layers, progressively extracting deep features from low-level edges and textures to high-level semantics, thus forming an abstract representation of the image content. Based on this, a Region Proposal Network (RPN) is used to automatically generate multiple candidate regions that may contain power fittings. Subsequently, each candidate region undergoes classification and bounding box regression: the classification task calculates the probability that the region belongs to "power fittings," while the regression task fine-tunes the position and size of the candidate region to obtain more accurate bounding box coordinates. Finally, the system determines all candidate regions with probability values ​​higher than a preset confidence threshold as valid detection results, confirming the presence of power fittings in the image, and saves the bounding box coordinate information corresponding to these valid regions for subsequent cropping.

[0075] In this embodiment, the object detection model can also be YOLOv8, Faster R-CNN, or DETR.

[0076] In this embodiment, cropping the inspection image to obtain an image of the power fitting includes: cropping the corresponding region from the original inspection image based on the bounding box coordinates output by the target detection, generating a local image of a single power fitting. The cropped image is then normalized (e.g., uniformly scaled to 512×512 pixels) while retaining the original aspect ratio information for subsequent area calculations. Optionally, basic image enhancement processing such as deblurring, denoising, or contrast enhancement is performed on the cropped image to improve the accuracy of subsequent analysis.

[0077] In some embodiments of this application, the power fitting image is preprocessed and color analyzed to obtain the surface color data of the power fitting. Specifically, the following steps are taken: the Retinex algorithm or white balance correction is used to correct the illumination of the power fitting image; image segmentation technology is applied to separate the power fitting from the background to obtain the power fitting body image; color features are extracted from the power fitting body image; the power fitting body image is converted from RGB space to CIELAB color space; dominant color clusters are extracted by K-means clustering; and the average L*, a*, and b* values ​​of each dominant color cluster and the proportion of pixel area they occupy are obtained to obtain color parameters and color area.

[0078] Specifically, this solution can eliminate the influence of interference factors such as uneven lighting and color deviation during image acquisition, accurately separate the body area of ​​the power fittings, and extract color features with physical meaning and perceptual consistency from them, thereby obtaining reliable surface color data that can be used for corrosion judgment.

[0079] Specifically, the cropped images of the power fittings undergo illumination correction using the Retinex algorithm or white balance correction method to reduce brightness unevenness or color distortion caused by changes in ambient lighting, making the images closer to the true colors observed by the human eye under standard lighting conditions. Subsequently, image segmentation techniques (such as edge-based, threshold-based, or semantic segmentation methods) are applied to accurately separate the power fitting body from the background, generating an image containing only the effective area of ​​the fitting, avoiding interference from background color in subsequent analysis. Based on this, color features are extracted from the fitting body image: first, its color representation is converted from the device-related RGB color space to the CIELAB color space, which is more consistent with human visual perception; then, the K-means clustering algorithm is used to cluster the color values ​​of all pixels, automatically identifying several dominant color clusters; finally, for each dominant color cluster, its average L* (brightness), a* (red-green axis), and b* (yellow-blue axis) values ​​are calculated as color parameters, and the proportion of pixels contained in this cluster to the total number of pixels on the fitting is calculated as the area proportion of the corresponding color. The color parameters and their corresponding area ratios obtained in this way constitute the surface color data of power fittings, which are used for subsequent corrosion identification and quantitative analysis.

[0080] In this embodiment, if the inspected image has local strong shadows or high dynamic range illumination, multi-scale Retinex (MSR / MSRCR) is preferred; if the overall color temperature of the image is significantly off but the illumination is relatively uniform, gray world white balance can be used; in a real system, an adaptive switching mechanism can be designed: the illumination complexity is judged based on the image gradient variance or HSV spatial saturation, and the correction strategy is automatically selected.

[0081] In this embodiment, illumination correction includes, but is not limited to, the Retinex algorithm, white balance algorithm, homomorphic filtering, or deep learning illumination decomposition network.

[0082] In some embodiments of this application, based on a preset power fitting database, the installation type of the power fitting is determined according to the shooting location, and the material parameters of the power fitting are determined based on the installation type. Specifically, the preset power fitting database includes several road segment information, each road segment information corresponding to the installation type, quantity, and time of the power fitting; the corresponding road segment information is determined according to the shooting location, the preset power fitting database is filtered, and the installation type of the power fitting corresponding to the shooting location is determined; the material parameters of the power fitting are determined according to the installation type, the material parameters including the metal composition of the fitting base, the surface coating type, the surface coating color, and the surface coating thickness.

[0083] Specifically, this solution retrieves the installation type of the power fittings from a pre-set database based on their specific installation location, and then determines the material parameters of the power fittings accordingly. First, the pre-set power fitting database contains information on different road sections, each recording details such as the installation type, quantity, and installation time of the power fittings. Once the location of a particular power fitting is captured, the corresponding road section information can be accurately determined by comparing it with the road section information in the database, thus filtering out the installation type of the power fitting that matches the captured location. This process ensures accurate identification of the type of power fitting used at a specific location, even in large-scale power grid systems. Next, once the installation type of the power fitting is determined, its specific material parameters can be further determined. These material parameters include, but are not limited to, the metal composition of the fitting base (such as aluminum, steel, etc.), the type of surface coating (such as zinc plating, epoxy resin coating, etc.), the color of the surface coating, and the thickness of the surface coating. This detailed information is crucial for analyzing the working condition of power fittings. For example, different metal compositions and surface coatings directly affect the corrosion resistance and service life of the fittings; while the color and thickness of the surface coating help determine whether its current protective condition is adequate and whether repair or replacement is necessary. Thus, accurate understanding of material parameters provides a scientific basis for subsequent maintenance work, ensuring the safe and stable operation of the power system.

[0084] In some embodiments of this application, the corrosion color of the power fitting is determined based on the material parameters of the power fitting. Specifically, a material-corrosion color mapping table is preset, which includes several metal materials. Each metal material is provided with a standard corrosion color range under standard atmospheric corrosion conditions. The metal composition of the fitting substrate is matched with the material-corrosion color mapping table to determine the standard corrosion color range corresponding to the matching of the metal composition of the fitting substrate with the metal material.

[0085] Specifically, this solution determines the range of colors that electrical fittings may exhibit when they rust, based on the material properties of the fittings, providing a scientific basis for subsequent image color analysis to determine whether corrosion has occurred.

[0086] Specifically, a material-rust color mapping table is pre-constructed, which includes various common metallic materials (such as carbon steel, aluminum alloys, and copper alloys). For each metallic material, under standard atmospheric corrosion conditions (such as the typical environment specified in ISO 9223), the standard rust color range corresponding to its typical rust products is defined. These color ranges are represented in the form of numerical intervals in a color space, for example, defining the upper and lower limits of L*, a*, and b* in the CIELAB color space. When the base metal composition of a certain electrical fitting is known, it is matched with the metallic material entries in the mapping table. Once a matching or closest metallic material entry is found, the standard rust color range corresponding to that material can be read from the mapping table. This range serves as a benchmark reference for rust identification, used for subsequent comparison with color parameters extracted from the actual image to determine whether there are color areas on the surface of the fitting that conform to rust characteristics.

[0087] In some embodiments of this application, the color parameters in the surface color data of the power fitting are compared with the rust color to determine whether the power fitting has rusted. Specifically, the color parameters are matched with the numerical values ​​of the surface coating color, and the color parameters that successfully match the numerical values ​​of the color parameters and the surface coating color are discarded to obtain suspected rust color parameters. The minimum color difference between the suspected rust color parameters and the standard rust color range is determined based on the suspected rust color parameters and the standard rust color range. The minimum color difference is compared with a preset color difference. If the minimum color difference is less than or equal to the preset color difference, the power fitting is determined to have rusted; if the minimum color difference is greater than the preset color difference, the power fitting is determined not to have rusted.

[0088] Specifically, this method can scientifically and accurately determine whether the surface of power fittings has rusted, by comparing the color parameters extracted from the image with the standard rust color range corresponding to the material, while excluding interference from the normal surface coating color of the power fittings. First, all color parameters extracted from the image of the power fitting (such as the L*, a*, and b* values ​​of each dominant color cluster) are matched with the known values ​​of the original surface coating color of the fitting. If a certain color parameter is consistent with the coating color within a preset tolerance range, it is considered to belong to the normal area without rust and is removed from the analysis, thus retaining the "suspected rust color parameters" that can only be caused by abnormal discoloration. Subsequently, for each suspected rust color parameter, the minimum color difference between it and the standard rust color range determined based on the material parameters is calculated—that is, the shortest color distance from the color parameter to the boundary of the rust color range, usually measured using color difference formulas such as CIEDE2000 or ΔE. Finally, the minimum color difference is compared with a preset color difference threshold: if the minimum color difference is less than or equal to the threshold, it means the color is close enough to the typical rust color and is determined to be rust; conversely, if the minimum color difference is greater than the threshold, the color is considered not to be a rust characteristic and is determined not to be rust. Through this series of logical judgments, normal coating aging, stains, or light reflection are effectively avoided from being misjudged as rust, improving the accuracy and reliability of rust identification.

[0089] In this embodiment, it is assumed that the standard corrosion color range is: L*: 40–60; a*: 10–25; b*: 15–30. The color parameters in the surface color data of the power fittings are compared with the corrosion colors to determine whether the power fittings have corroded. Alternatively, the average color value can be directly compared with the boundaries of the standard corrosion color range, i.e., checking whether L* falls between 40 and 60, a* between 10 and 25, and b* between 15 and 30.

[0090] In some embodiments of this application, the real-time corrosion degree of the power fitting is determined based on the color area in the surface color data of the power fitting. Specifically, the color area corresponding to each suspected corrosion color parameter is set as the suspected corrosion color area; the sum of the suspected corrosion color areas is calculated, the difference between the sum of the suspected corrosion color areas and the total surface area of ​​the power fitting is determined, and the difference is set as the real-time corrosion degree of the power fitting.

[0091] Specifically, this solution can quantify the degree of corrosion of electrical fittings after confirming that corrosion has occurred, and obtain a measurable and comparable "real-time corrosion degree" index to provide data support for subsequent condition assessment and maintenance decisions.

[0092] Specifically, after completing the corrosion assessment, the system defines the area corresponding to all color parameters identified as potentially corroded as "potentially corroded color areas." These areas are typically expressed in terms of pixel count or normalized ratio. Subsequently, all potentially corroded color areas are summed to obtain the total potentially corroded area. Next, the total surface area of ​​the power fitting is obtained (which can be calculated from the total number of pixels in the fitting's body area in the image or by combining a geometric model), and the difference between the total surface area and the non-corroded area is calculated, or the sum of the potentially corroded color areas is directly considered as the area of ​​the corroded portion. According to the description in the claim, this difference is set as the "real-time corrosion degree" of the power fitting, used to characterize the degree to which the fitting surface is covered by corrosion at the current inspection moment. This value can be further compared with the predicted corrosion degree to determine whether the corrosion development is abnormal and to support the classification of corrosion severity levels.

[0093] In some embodiments of this application, meteorological data of the corresponding location of the power fitting is obtained based on the shooting location. The corrosion reaction of the power fitting is predicted based on the meteorological data and the material parameters of the power fitting to obtain the predicted corrosion degree. Specifically, the installation time of the power fitting is determined according to a preset power fitting database, and meteorological data of the corresponding time and location of the power fitting is obtained based on the installation time and shooting location. The meteorological data includes rainfall, temperature and humidity, precipitation frequency, salt spray concentration, etc. Concentration; Based on a pre-built corrosion prediction model, the predicted corrosion degree of power fittings is determined according to the material parameters and meteorological data of the power fittings.

[0094] Specifically, this solution can combine the actual service environment and material properties of power fittings to scientifically predict their corrosion development trend under current usage conditions, thereby obtaining a theoretical "predicted corrosion degree" which can be compared with the actual test results to determine whether the corrosion is abnormal or accelerated.

[0095] Specifically, the installation time of the electrical fitting is retrieved from a pre-set database of electrical fittings, and combined with its shooting location (i.e., geographical coordinates), historical meteorological data experienced at that location from the date of installation to the current inspection time is obtained. This meteorological data includes key environmental factors that significantly affect metal corrosion, such as cumulative rainfall, average temperature and humidity, precipitation frequency, salt spray concentration in the air, and sulfur dioxide (SO2). The meteorological data, along with the determined material parameters of the power fittings (such as base metal composition, coating type, and thickness), are then input into a pre-built corrosion prediction model. This model, based on corrosion science principles or trained using historical corrosion samples, comprehensively considers the interaction between material corrosion resistance and environmental corrosion, simulating the accumulation process of corrosion on the fitting surface under specific service conditions. It ultimately outputs a value reflecting the theoretical degree of corrosion, i.e., the "predicted corrosion degree." This predicted value represents the corrosion level that should be achieved under normal environment-material coupling, providing a benchmark for subsequent assessment of whether the actual corrosion state deviates from expectations.

[0096] In some embodiments of this application, the severity level of corrosion on the surface of power fittings is determined based on the real-time corrosion degree and the predicted corrosion degree. Specifically, the real-time corrosion degree and the predicted corrosion degree are compared. If the real-time corrosion degree is less than or equal to the predicted corrosion degree, the corrosion on the surface of the power fittings is determined to be not severe. If the real-time corrosion degree is greater than the predicted corrosion degree, the corrosion degree difference between the real-time corrosion degree and the predicted corrosion degree is determined, and the severity level of corrosion on the surface of the power fittings is determined based on the corrosion degree difference. A first corrosion degree difference and a second corrosion degree difference are preset, wherein the first corrosion degree difference is less than the second corrosion degree difference. The severity level of corrosion on the surface of the power fittings is determined based on the corrosion degree difference and the predicted corrosion degree. The relationship between the first rust degree difference and the second rust degree difference determines the rust severity level of the power fitting surface; if the rust degree difference is less than the first rust degree difference, the rust severity level of the power fitting surface is set to Level 1 severe; if the rust degree difference is greater than or equal to the first rust degree difference and less than the second rust degree difference, the rust severity level of the power fitting surface is set to Level 2 severe; if the rust degree difference is greater than or equal to the second rust degree difference, the rust severity level of the power fitting surface is set to Level 3 severe, wherein the severity of Level 1, Level 2, and Level 3 severe increases sequentially.

[0097] Specifically, this solution classifies and assesses the corrosion status by comparing the deviation between the actual corrosion level and the theoretically predicted corrosion level of power fittings, thereby providing a clear and actionable basis for determining the severity level for operation and maintenance decisions.

[0098] Specifically, the system first compares the real-time corrosion rate obtained through image analysis with the predicted corrosion rate calculated based on material parameters and meteorological data. If the real-time corrosion rate is less than or equal to the predicted corrosion rate, it indicates that the corrosion development of the hardware is normal or relatively slow, without abnormal accelerated corrosion, and therefore the corrosion is not severe, requiring no emergency intervention. If the real-time corrosion rate is greater than the predicted corrosion rate, it indicates that the actual corrosion degree exceeds the expected level under the coupling effect of the environment and materials, and there may be a localized harsh microenvironment, coating failure, or other abnormal factors, requiring further risk assessment. In this case, the system calculates the difference between the two, i.e., the "corrosion rate difference". Two thresholds are preset: a first corrosion rate difference and a second corrosion rate difference, and the first corrosion rate difference is less than the second corrosion rate difference. Based on the relationship between the difference in corrosion severity and these two thresholds, corrosion severity is classified into three levels: Level 1 severe when the difference is less than the first corrosion severity difference; Level 2 severe when the difference is greater than or equal to the first corrosion severity difference but less than the second corrosion severity difference; and Level 3 severe when the difference is greater than or equal to the second corrosion severity difference. These three levels increase in severity sequentially and correspond to different maintenance response strategies, such as regular observation, planned maintenance, or immediate replacement, thereby achieving refined management and tiered handling of corrosion risks in power fittings.

[0099] See Figure 2 As shown, this invention also discloses a machine vision-based intelligent surface monitoring system for power fittings, used to apply the above-mentioned machine vision-based intelligent surface monitoring method for power fittings, comprising:

[0100] The power fittings image acquisition module is configured to acquire inspection images of transmission lines, perform target detection on the inspection images, determine whether the inspection images contain power fittings, and if they contain power fittings, crop the inspection images to obtain power fittings images.

[0101] The power fitting color analysis module is configured to preprocess and perform color analysis on the power fitting image to obtain power fitting surface color data, which includes color parameters and color area.

[0102] The power fitting material determination module is configured to acquire the shooting location of the inspection image, determine the installation type of the power fitting based on the shooting location according to the preset power fitting database, and determine the material parameters of the power fitting based on the installation type.

[0103] The power fitting corrosion determination module is configured to determine the corrosion color of the power fitting based on the material parameters of the power fitting, compare the color parameters in the surface color data of the power fitting with the corrosion color to determine whether the power fitting has rusted, and if corrosion is determined, determine the real-time corrosion degree of the power fitting based on the color area in the surface color data of the power fitting.

[0104] The corrosion severity judgment module is configured to acquire meteorological data of the corresponding location of the power fitting based on the shooting location, predict the corrosion reaction of the power fitting based on the meteorological data and the material parameters of the power fitting, obtain the predicted corrosion degree of the power fitting, and judge the corrosion severity level of the power fitting surface based on the real-time corrosion degree and the predicted corrosion degree.

[0105] This invention utilizes an image acquisition module for power fittings to perform precise target detection and cropping of inspection images, effectively focusing on the power fittings themselves, eliminating background interference, and providing high-quality image data for subsequent analysis. The color analysis module for power fittings, through preprocessing and color analysis, not only extracts accurate color parameters but also quantifies color area, providing objective quantitative evidence for corrosion assessment. The material determination module for power fittings, combining the shooting location with a preset database, accurately identifies the installation type and corresponding material parameters of the power fittings, making corrosion color assessment more targeted and scientific. The corrosion determination module for power fittings compares actual color parameters with corrosion colors determined based on materials and combines color area to determine real-time corrosion degree, achieving accurate identification and quantitative assessment of corrosion conditions. The corrosion severity assessment module innovatively incorporates meteorological data and combines it with material parameters to predict corrosion reactions. By comparing real-time corrosion degree with predicted corrosion degree, it can more comprehensively and dynamically assess the corrosion severity level, providing more scientific and forward-looking decision support for the maintenance and replacement of power fittings, helping to promptly identify potential risks and ensure the safe and stable operation of transmission lines.

[0106] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program goods. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program goods embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0107] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program goods according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0108] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0109] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0110] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A machine vision-based intelligent monitoring method for the surface of electrical fittings, characterized in that, include: Obtain inspection images of transmission lines, perform target detection on the inspection images to determine whether the inspection images contain power fittings, and if they contain power fittings, crop the inspection images to obtain power fitting images. The power fitting image is preprocessed and color analyzed to obtain the power fitting surface color data, which includes color parameters and color area. The system acquires the shooting location of the inspection image, determines the installation type of the power fitting based on the preset power fitting database, and determines the material parameters of the power fitting based on the installation type. The corrosion color of the power fitting is determined based on the material parameters of the power fitting. The color parameters in the surface color data of the power fitting are compared with the corrosion color to determine whether the power fitting has rusted. If corrosion is determined, the real-time corrosion degree of the power fitting is determined based on the color area in the surface color data of the power fitting. Meteorological data of the corresponding location of the power fittings are obtained based on the shooting location. The corrosion reaction of the power fittings is predicted based on the meteorological data and the material parameters of the power fittings. The predicted corrosion degree of the power fittings is obtained. The severity level of corrosion on the surface of the power fittings is determined based on the real-time corrosion degree and the predicted corrosion degree.

2. The intelligent surface monitoring method for power fittings based on machine vision according to claim 1, characterized in that, Target detection is performed on the inspection image to determine whether the inspection image contains power fittings, specifically: The inspection images are input into a pre-trained power fittings target detection model; Feature extraction is performed on the inspection image. Deep semantic features of the inspection image are obtained through convolutional layers and pooling layers. Candidate power fitting regions are generated using a region generation network. The candidate power fitting regions are classified and bounding box regressed to obtain the probability value of each candidate power fitting region belonging to power fitting and the corresponding precise bounding box coordinates. When the probability value of a candidate power fitting area is greater than the preset confidence threshold, it is determined that the inspection image contains power fittings, and the coordinate information of all bounding boxes that meet the conditions is recorded.

3. The intelligent surface monitoring method for power fittings based on machine vision according to claim 1, characterized in that, The images of the power fittings are preprocessed and subjected to color analysis to obtain the surface color data of the power fittings, specifically: Illumination correction is performed on images of power fittings using the Retinex algorithm or white balance correction. Image segmentation technology is applied to separate the power fittings from the background to obtain the image of the power fitting itself; Color features are extracted from the image of the power fitting body. The image of the power fitting body is converted from RGB space to CIELAB color space. Dominant color clusters are extracted by K-means clustering. The average L*, a*, b* values ​​of each dominant color cluster and the proportion of the pixel area occupied are obtained to obtain color parameters and color area.

4. The intelligent surface monitoring method for power fittings based on machine vision according to claim 1, characterized in that, Based on a pre-set database of power fittings, the installation type of the power fitting is determined according to the shooting location, and the material parameters of the power fitting are determined based on the installation type, specifically as follows: The preset power fittings database includes information on several road segments, each segment corresponding to the installation type, quantity, and installation time of the power fittings. Based on the shooting location, the corresponding road segment information is determined, and the preset power fitting database is filtered to determine the power fitting installation type corresponding to the shooting location; The material parameters of the power fittings are determined according to the installation type of the power fittings. The material parameters include the metal composition of the fitting base, the type of surface coating, the color of the surface coating, and the thickness of the surface coating.

5. The intelligent surface monitoring method for power fittings based on machine vision according to claim 4, characterized in that, The corrosion color of the power fittings is determined based on their material parameters, specifically as follows: A material-rust color mapping table is preset, which includes several metal materials, and each metal material is given a standard rust color range under standard atmospheric corrosion conditions. The metal composition of the hardware substrate is matched with the material-rust color mapping table to determine the standard rust color range corresponding to the match between the metal composition of the hardware substrate and the metal material.

6. The intelligent surface monitoring method for power fittings based on machine vision according to claim 5, characterized in that, The color parameters in the surface color data of the power fittings are compared with the rust color to determine whether the power fittings have rusted. Specifically: The color parameters are matched with the values ​​of the surface coating color. Color parameters that successfully match the values ​​of the surface coating color are discarded to obtain suspected rust color parameters. The minimum color difference between the suspected rust color parameter and the standard rust color range is determined based on the suspected rust color parameter and the standard rust color range. The minimum color difference is compared with the preset color difference. If the minimum color difference is less than or equal to the preset color difference, it is determined that the power fitting has rusted; if the minimum color difference is greater than the preset color difference, it is determined that the power fitting has not rusted.

7. The intelligent surface monitoring method for power fittings based on machine vision according to claim 6, characterized in that, The real-time corrosion degree of the power fittings is determined based on the color area in the surface color data of the power fittings. Specifically, the color area corresponding to each suspected corrosion color parameter is set as the suspected corrosion color area. Calculate the sum of the suspected rust color areas, determine the difference between the sum of the suspected rust color areas and the total surface area of ​​the power fittings, and set the difference as the real-time rust degree of the power fittings.

8. The intelligent surface monitoring method for power fittings based on machine vision according to claim 1, characterized in that, Meteorological data of the corresponding location of the power fittings are obtained based on the shooting location. The corrosion reaction of the power fittings is predicted based on the meteorological data and the material parameters of the power fittings, and the predicted corrosion degree of the power fittings is obtained, specifically as follows: The installation time of the power fittings is determined based on a pre-set power fittings database. Meteorological data corresponding to the installation time and location of the power fittings are then obtained based on the installation time and shooting location. This meteorological data includes rainfall, temperature and humidity, precipitation frequency, and salt spray concentration. concentration; Based on a pre-built corrosion prediction model, the predicted corrosion degree of power fittings is determined according to the material parameters and meteorological data.

9. The intelligent surface monitoring method for power fittings based on machine vision according to claim 1, characterized in that, The severity of corrosion on the surface of power fittings is determined based on the real-time corrosion degree and the predicted corrosion degree, specifically as follows: The real-time corrosion degree is compared with the predicted corrosion degree. If the real-time corrosion degree is less than or equal to the predicted corrosion degree, it is determined that the corrosion on the surface of the power fittings is not serious. If the real-time corrosion degree is greater than the predicted corrosion degree, then the corrosion degree difference between the real-time corrosion degree and the predicted corrosion degree is determined, and the corrosion severity level of the power fitting surface is judged based on the corrosion degree difference. A first rust degree difference and a second rust degree difference are preset, wherein the first rust degree difference is less than the second rust degree difference; The corrosion severity level of the power fitting surface is set according to the relationship between the corrosion degree difference and the first corrosion degree difference and the second corrosion degree difference; If the corrosion degree difference is less than the first corrosion degree difference, then the corrosion severity level of the power fitting surface is set to Level 1 severe level. If the corrosion degree difference is greater than or equal to the first corrosion degree difference, and the corrosion degree difference is less than the second corrosion degree difference, then the corrosion severity level of the power fitting surface is set to Level 2 severe level. If the corrosion degree difference is greater than or equal to the second corrosion degree difference, the corrosion severity level of the power fitting surface is set to Level 3, wherein the severity of Level 1, Level 2, and Level 3 increases sequentially.

10. A machine vision-based intelligent surface monitoring system for power fittings, used to apply the machine vision-based intelligent surface monitoring method for power fittings as described in any one of claims 1-9, characterized in that, include: The power fittings image acquisition module is configured to acquire inspection images of transmission lines, perform target detection on the inspection images, determine whether the inspection images contain power fittings, and if they contain power fittings, crop the inspection images to obtain power fittings images. The power fitting color analysis module is configured to preprocess and perform color analysis on the power fitting image to obtain power fitting surface color data, which includes color parameters and color area. The power fitting material determination module is configured to acquire the shooting location of the inspection image, determine the installation type of the power fitting based on the shooting location according to the preset power fitting database, and determine the material parameters of the power fitting based on the installation type. The power fitting corrosion determination module is configured to determine the corrosion color of the power fitting based on the material parameters of the power fitting, compare the color parameters in the surface color data of the power fitting with the corrosion color to determine whether the power fitting has rusted, and if corrosion is determined, determine the real-time corrosion degree of the power fitting based on the color area in the surface color data of the power fitting. The corrosion severity judgment module is configured to acquire meteorological data of the corresponding location of the power fitting based on the shooting location, predict the corrosion reaction of the power fitting based on the meteorological data and the material parameters of the power fitting, obtain the predicted corrosion degree of the power fitting, and judge the corrosion severity level of the power fitting surface based on the real-time corrosion degree and the predicted corrosion degree.

Images