An image analysis-based insulation boot installation monitoring system and insulation boot

Abstract

This invention belongs to the field of power facility monitoring technology, specifically an image analysis-based monitoring system for the installation of insulating protective covers and the insulating protective cover itself. The system integrates image acquisition, preprocessing, algorithm analysis, color recognition, end-edge collaboration modules, and a sample library. The protective cover uses a magnetic self-locking component and a thermochromic coating. The system improves image quality through illumination compensation and multi-scale filtering, extracts topological features at the opening and identifies pixel gaps to determine the physical closure status, and automatically captures temperature rise signals using color offset. This application solves the problems of manual inspection deviation and environmental interference, and realizes closed-loop monitoring from installation accuracy recognition to color change early warning during operation.

Description

Technical Field

[0001] This invention belongs to the field of power facility monitoring technology, specifically an image analysis-based monitoring system for the installation of insulating protective covers and the insulating protective cover itself. Background Technology

[0002] In the operation and maintenance of power distribution networks, power equipment with voltage levels of 10 kV and below is widely distributed and operates in a complex environment. As the core line of defense for equipment insulation protection, the installation quality and operating status of insulating protective covers are directly related to the reliability of the power grid and public safety. Traditional manual inspections are affected by factors such as terrain, perspective, and distance, resulting in subjective bias, low efficiency, missed inspections, and misjudgments. This can easily lead to safety hazards such as loose or detached protective covers.

[0003] With the digital transformation of the power grid, image analysis and automated monitoring technology has been introduced into the inspection process. However, the existing general image processing algorithms are not semantically coupled with the monitoring of special power grid fittings. Environmental interference such as strong outdoor light, rain, fog, and vegetation leads to uncertain image features of the protective cover. Traditional preprocessing technology is difficult to balance detail and noise reduction, and the monitoring model has weak generalization ability and cannot adapt to diverse inspection scenarios with different fitting topologies and different terrain backgrounds, making it difficult to meet the needs of precise operation and maintenance.

[0004] Existing monitoring systems suffer from lag in monitoring operational status evolution, making it impossible to effectively capture early warning signals of thermal discoloration of insulating protective covers. Outdoor ambient light distortion can easily mask the true temperature rise and discoloration, leading to false warnings. Furthermore, the existing architecture focuses on centralized cloud processing, making it difficult for high-precision models to run in real time at the edge and achieve end-edge collaboration. This results in a logical disconnect between installation monitoring and fault early warning, making it impossible to build a closed-loop monitoring system covering the entire lifecycle of the protective cover.

[0005] Therefore, the present invention provides an image analysis-based monitoring system for the installation of an insulating protective cover and an insulating protective cover. Summary of the Invention

[0006] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0007] The technical solution adopted by this invention to solve its technical problem is as follows: An image analysis-based insulation protective cover installation monitoring system, comprising an image acquisition module, an image preprocessing module, an image analysis algorithm module, a color recognition model module, an end-edge collaborative interaction module, and an insulation protective cover image sample library. Through signal flow and logical coordination between the above modules, the system constructs a highly deterministic monitoring link. The image acquisition module is used to acquire the original optical information of the insulation protective cover at the power distribution node and convert this information into a digital image signal stream to be processed.

[0008] Preferably, the image preprocessing module performs illumination compensation processing on the original image. Through adaptive histogram equalization logic, it remaps the pixel brightness distribution in different exposure areas to eliminate local feature loss caused by uneven light and shadow distribution. The image preprocessing module also performs noise suppression processing based on multi-scale guided filtering. This process utilizes a spatial linear model of the guided image to filter out high-frequency random noise while performing edge preservation operations on high-frequency details reflecting the mechanical edges of the protective shield, ensuring that the geometric contours of the protective shield have clear boundary gradients at the pixel level. The image preprocessing module also integrates motion deblurring functionality, using a spatially variable point spread function to perform deconvolution correction on the image displacement generated during the inspection process, thereby providing high-fidelity feature vector input for subsequent feature extraction.

[0009] Preferably, the image analysis algorithm module is the core logic unit for achieving refined installation identification. This module, through the development of an image analysis algorithm adapted to the protective cover, accurately identifies physical states such as whether the installation is in place and whether the magnetic self-locking is secure. At the logic execution level, the image analysis algorithm module first extracts the topological features of the opening of the insulating protective cover. These topological features include the curvature of the opening edge in the pixel coordinate system, the gap width, and the shadow distribution pattern. The module determines the physical closure state of the magnetic self-locking component by identifying the pixel gap width. When the protective cover is in a fully closed self-locking state, the pixel gap at its opening edge approaches zero, and the edge features exhibit a continuous and smooth closed trajectory. If the installation is not in place, the image analysis algorithm module will capture discrete shadow transition points and discontinuous pixel holes. This module compares the real-time extracted spatial geometric features with the preset installation standard template using feature vectors, outputting a definitive judgment result on the installation's adequacy, thereby significantly improving the detection accuracy.

[0010] Preferably, the insulating protective cover undergoes deep functional modification in its physical structure. Its opening edge integrates a magnetic self-locking component. This component utilizes the physical logic of opposite magnetic poles attracting to generate an automatic attraction force when the installation and covering action approaches the closing point, ensuring effective locking of the mechanical structure. Simultaneously, the surface of the insulating protective cover is coated with a thermochromic coating. This thermochromic coating exhibits a distinct color-changing warning characteristic at 60 to 80 degrees Celsius. When the internal hardware of the protective cover exceeds a preset threshold due to temperature rise, the thermochromic coating undergoes a reflectance spectral shift, producing a physical color change from the initial color to a warning color, providing a physical trigger logic for subsequent fault detection.

[0011] Preferably, the color recognition model module executes an automatic thermochromic capture program. This module integrates the thermochromic characteristics of the protective cover at 60 to 80 degrees Celsius to construct a dedicated image color recognition model. The color recognition model module first performs color correction in the HSV color space, eliminating false features caused by ambient light color shifts in the morning or evening by extracting the chromaticity components of preset reference objects in the environment. Subsequently, the module extracts the real-time color feature components of the insulating protective cover surface and automatically captures and quantifies abnormal temperature-induced color changes by calculating the offset of these components relative to a preset color threshold. This quantitative analysis model transforms the physical color change characterization into digital fault semantics, achieving the integration of installation monitoring and fault early warning.

[0012] Preferably, the edge-to-edge collaborative interaction module is responsible for constructing the communication and computing closed loop of the lightweight monitoring system. To lower the application threshold and ensure compatibility with existing power grid inspection equipment, this module executes edge-to-edge collaborative logic. In terms of computing logic allocation, the inspection terminal, acting as the edge side, performs preliminary analysis of the lightweight model. Through parameter pruning and knowledge distillation techniques, a lightweight algorithm is obtained to quickly predict images at the edge side. When a suspected abnormal state is detected, the edge side transmits feature packets containing local details of the abnormality back to the central management platform via a high-priority link for in-depth verification. This architecture reduces the continuous dependence on mobile communication bandwidth, effectively alleviates the processing pressure on the cloud, and realizes edge-to-edge collaboration in image acquisition, analysis, and early warning.

[0013] Preferably, the protective cover image sample library is used to optimize the model's generalization ability. This sample library establishes a set of protective cover status images containing different hardware topologies, terrain backgrounds, and meteorological conditions. The sample library utilizes data augmentation technology and generative adversarial networks to simulate imaging effects under different viewing angles, spatial topological interference, and observation distances. This ensures that the monitoring model maintains stable recognition rates when facing different hardware scenarios such as transformer terminals, circuit breakers, and surge arresters, as well as different terrain backgrounds such as mountainous areas, forest areas, and urban areas. This allows it to adapt to protective cover monitoring scenarios with different hardware and terrains. Through the continuous evolution of the sample library, the system achieves seamless integration with the power grid intelligent operation and maintenance system.

[0014] Preferably, when performing monitoring tasks, the present invention first activates the image acquisition module to capture visual information at the target hardware location. The image preprocessing module then activates, improving the image signal-to-noise ratio through illumination compensation and noise suppression logic. The image analysis algorithm module then performs a topological scan of the magnetic self-locking points of the insulating protective cover, determining the installation's proper placement through pixel-level gap judgment. If the installation is deemed satisfactory, the system automatically switches to the monitoring mode. The color recognition model module continuously monitors the color vector of the thermochromic coating. Once a color shift exceeds the warning threshold, the system triggers a fault alarm through the edge-to-edge collaborative interaction module. Monitoring data throughout the entire lifecycle is synchronized in real-time to the intelligent operation and maintenance system, providing data support for subsequent preventative maintenance.

[0015] Preferably, through improvements to the hardware structure (magnetic self-locking and thermochromic properties) and reconstruction of the software algorithm (semantic perception preprocessing, deep recognition model, and edge-end collaborative architecture), a logically complete monitoring system has been constructed.

[0016] Preferably, the insulation protective cover installation monitoring system of the present invention exhibits extremely high synergy in physical connection and logical transmission. The semantic perception extraction logic ensures that the hardware can understand the spatial topological relationship between the fittings and the protective cover, while the active data flow mechanism guarantees millisecond-level interaction of monitoring results between the inspection terminal and the central platform. Through in-depth characterization of the self-locking characteristics of the magnetic attraction structure, the system transforms the physical closed state into a closed loop determination in topological graph theory, greatly enhancing the technical certainty under legal support. The design of each functional module directly addresses the defects mentioned in the background technology. Through decoupling and reconstruction at the mechanism level, intelligent control of the entire process of power grid equipment protection is ultimately achieved.

[0017] Preferably, the image preprocessing module of the present invention employs the dark channel prior principle to perform dehazing when dealing with rain and fog obscuration, restoring the image's contrast and saturation. When dealing with vegetation background interference, the module uses color space segmentation logic to logically remove green vegetation noise from the background, highlighting the shape features of the protective cover's main body. This series of procedural details supports a higher-level overview of preprocessing technology, ensuring the robustness of the monitoring logic under actual working conditions.

[0018] Preferably, the insulating protective cover of the present invention employs a reversible microcapsule color-developing mechanism in its thermochromic coating, ensuring the persistence and reliability of the warning signal. The magnetic self-locking component uses high-performance rare-earth permanent magnet materials, guaranteeing locking strength under long-term vibration. These physical characteristics form a strict one-to-one correspondence with the color recognition and topology analysis in the software algorithm, ensuring the system's absolute perception of key states such as proper installation, secure self-locking, and color-changing warning.

[0019] The beneficial effects of this invention are as follows:

[0020] 1. The present invention discloses an image analysis-based monitoring system for the installation of an insulating protective cover and the insulating protective cover itself. Through an image analysis algorithm module, the system performs pixel-level scanning and feature vector comparison of the topological features (including edge curvature, gap width, shadow distribution, etc.) at the opening of the insulating protective cover. This transforms the traditional qualitative judgment relying on manual visual identification into a quantitative analysis based on the closed trajectory of pixel gaps. When the magnetic self-locking component is fully closed, the system can accurately identify the closed loop features of continuous, smooth opening edges with gaps approaching zero. If the installation is not in place, the system can instantly capture discrete shadow transition points and discontinuous pixel holes. This mechanism fundamentally eliminates subjective judgment bias, significantly improving the accuracy of installation quality inspection and the technical certainty from a legal perspective.

[0021] 2. The image analysis-based monitoring system for the installation of insulating protective covers and the insulating protective cover described in this invention eliminate local feature loss caused by uneven illumination through adaptive histogram equalization; utilizes multi-scale guided filtering to accurately preserve the edge contour of the protective cover while filtering out high-frequency random noise; employs spatially variable point spread function inverse convolution to correct motion blur; introduces dark channel prior defogging processing for rain and fog scenes; and effectively removes vegetation background interference using color space segmentation logic. These preprocessing techniques work synergistically to transform the original image under strong outdoor interference conditions into a feature vector input with high signal-to-noise ratio and high edge sharpness, ensuring that the subsequent recognition model maintains stable detection performance under complex working conditions.

[0022] 3. The image analysis-based installation monitoring system and insulating protective cover of this invention integrate a thermochromic coating with color-changing characteristics at 60 to 80 degrees Celsius into the physical structure and construct a dedicated image color recognition model. This model performs color correction in the HSV color space, eliminates the color shift of dawn and dusk by extracting the chromaticity components of environmental reference objects, and accurately captures the real-time color feature components of the coating surface, calculating its offset relative to a preset color threshold. This automatically quantifies the physical color change characterization into digital fault semantics. This design integrates installation monitoring and operational temperature rise early warning into the same system, overcoming the functional shortcomings of traditional methods that cannot detect potential thermal hazards in equipment in real time.

[0023] 4. The image analysis-based insulation cover installation monitoring system and insulation cover described in this invention utilize an edge-end collaborative interaction module to deploy a lightweight algorithm obtained through parameter pruning and knowledge distillation at the inspection terminal (edge ​​side), enabling rapid image prediction. Only when a suspected anomaly is detected is a feature packet containing local details of the anomaly sent back to the central platform for deep verification. This architecture significantly reduces the continuous occupation of mobile communication bandwidth, effectively alleviating cloud processing pressure, and avoids the need to deploy expensive and corroded online sensors at power distribution nodes. This lightweight and low-threshold design allows for large-scale adaptation to existing power grid inspection equipment, significantly reducing system construction and subsequent operation and maintenance costs.

[0024] 5. The present invention discloses an image analysis-based monitoring system for the installation of insulating protective covers and the insulating protective cover itself. This invention establishes a sample image library of protective cover status images containing different hardware topologies (transformer terminals, circuit breakers, surge arresters, etc.), different terrain backgrounds (mountainous areas, forest areas, urban areas, etc.), and different meteorological conditions. It utilizes data augmentation technology and generative adversarial networks to simulate visual occlusion, spatial topological interference, and imaging effects at different observation distances. Through continuous iterative optimization of the sample library, the monitoring model maintains stable recognition rates even when facing various hardware types and complex backgrounds. This effectively solves the performance degradation problem of traditional algorithms due to scene migration and achieves seamless integration with the smart grid operation and maintenance system.

[0025] 6. The image analysis-based installation monitoring system and insulating protective cover described in this invention provide a clear physical closure judgment basis for the image analysis algorithm through a magnetic self-locking component, and provide controllable optical response characteristics for the color recognition model through a thermochromic coating, forming a one-to-one correspondence between physical structure, visual features, and algorithm judgment. In the monitoring process, the system first performs an installation in place judgment. After the judgment is passed, it automatically switches to the operation monitoring mode, continuously tracking the coating color shift. Once the warning threshold is triggered, a fault alarm is pushed in real time through the edge-end collaboration module. The monitoring data throughout the entire life cycle is synchronized to the intelligent operation and maintenance system, forming a complete closed loop from installation acceptance to operation warning, providing reliable data support for preventive maintenance. Attached Figure Description

[0026] The invention will now be further described with reference to the accompanying drawings.

[0027] Figure 1 This is a structural block diagram of an image analysis-based monitoring system for the installation of an insulating protective cover, as described in this invention.

[0028] Figure 2 This is a perspective view of an insulating protective cover based on image analysis according to the present invention;

[0029] Figure 3This is a schematic diagram of the thermochromic coating in this invention.

[0030] In the picture: 1. Main body of the protective cover; 2. Magnetic self-locking component; 3. Thermochromic coating. Detailed Implementation

[0031] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0032] like Figure 1 As shown in the embodiment of the present invention, an image analysis-based insulation protective cover installation monitoring system includes an image acquisition module, an image preprocessing module, an image analysis algorithm module, a color recognition model module, an edge-to-edge collaborative interaction module, and an insulation protective cover image sample library. These modules are not isolated but logically connected through a standardized data bus and specific protocols, jointly supporting an automated architecture for monitoring the semantics of special hardware. During the initialization and operation phase of the system, the image acquisition module, as the sensing front end, uses a high-resolution charge-coupled device or complementary metal-oxide-semiconductor sensor configured on the inspection terminal or UAV mounting platform to capture the raw optical information of the insulation protective cover at the target power distribution node. The capture process adjusts the shutter speed and sensitivity to ensure that a digital image signal stream reflecting the physical boundary of the protective cover can be acquired under various lighting conditions. The signal stream is then pushed to the image preprocessing module in real time for preprocessing.

[0033] To address the uncertainty of feature vectors caused by drastic changes in light intensity, rain and fog obstruction, and complex vegetation backgrounds in outdoor environments, the image preprocessing module of this invention executes a multi-level depth-optimized preprocessing procedure. Furthermore, the module first performs illumination compensation processing on the original image. In its execution logic, the module employs adaptive histogram equalization logic. This logic does not simply stretch the entire image to grayscale; instead, it divides the digital image into several non-overlapping sub-regions. For each sub-region, it independently calculates its cumulative distribution function based on the pixel grayscale distribution characteristics. Through a contrast-limited mechanism, it prevents excessive amplification of background noise during the enhancement process, thereby achieving a dynamic balance mapping between bright and dark areas at the pixel level. This processing eliminates the loss of local features caused by uneven light and shadow distribution, ensuring that the mechanical boundary line of the insulating protective cover remains clearly visible even in complex backlighting environments.

[0034] The image preprocessing module performs noise suppression based on multi-scale guided filtering. The core of this process lies in utilizing the spatial linear model of the guided image. Taking the image to be processed as input, it obtains the output image through local linear transformation. Under the multi-scale architecture, the module can identify and filter out high-frequency random noise within different receptive fields. Due to the excellent edge-preserving properties of guided filtering, it can smooth pixels in uniform regions while performing precise gradient preservation operations on high-frequency details reflecting the mechanical edges of the protective shield. This ensures that the geometric contours of the insulating protective shield have extremely high boundary gradients at the pixel level, laying a solid foundation for subsequent topological feature extraction.

[0035] Furthermore, to address the unavoidable equipment vibration issue during inspection operations, the image preprocessing module integrates motion deblurring functionality. This function utilizes a spatially variable point diffusion function to model image trailing caused by relative displacement during acquisition. It then performs deconvolution correction using Wiener filtering or the Richardson-Lucy iterative algorithm, refocusing the diffused energy back to the original pixel locations. Additionally, for rainy and foggy weather common in power distribution networks, the image preprocessing module performs dehazing using the dark channel prior principle. By estimating the global atmospheric light intensity and transmittance map, it recovers a fog-free image with extremely high contrast and saturation from the damaged image, ensuring high fidelity of the feature vectors. When dealing with vegetation background interference, the module introduces color space segmentation logic, using threshold filtering of the green component in the HSV space to logically remove leaf noise from the background, achieving significant background decoupling of the insulating protective cover body 1 in the feature space.

[0036] The image analysis algorithm module described in this invention serves as the core logic unit for refined installation identification. Its development concept lies in transforming the macroscopic physical installation state into a microscopic topological graph theory determination. At the logic execution level, this module first extracts topological features from the preprocessed image. These topological features are key attributes describing the physical form and connection relationships of the insulating protective cover, specifically including the curvature change of the cover opening edge in the pixel coordinate system, the gap width distribution, and the shadow features generated by incomplete closure. Furthermore, the image analysis algorithm module accurately determines the physical closure state of the magnetic self-locking component 2 by identifying the pixel gap width. In a specific engineering implementation, the module uses an edge detection operator to locate the boundary line between the two half-shells of the protective cover and calculates the Euclidean distance between the edge pixels on both sides.

[0037] When the insulating protective cover is in a fully closed, self-locking state, the pixel gap at its opening edge approaches zero in the algorithm's interpretation, and the edge features exhibit a continuous and smooth closed trajectory in the topological space, without any logical breakpoints. Conversely, if the installation is not up to standard, resulting in incomplete installation, the image analysis algorithm module will capture discrete shadow transition points caused by incomplete magnetic pole engagement and identify discontinuous pixel holes in the image binarization mask. The module performs feature vector comparison by extracting these spatial geometric features in real time with the installation standard templates stored in the protective cover image sample library. The comparison process uses a weighted cross-entropy loss function for similarity measurement and outputs a deterministic judgment result on the installation's completion. This pixel-level quantitative discrimination logic elevates the vague qualitative judgment in traditional manual inspection to legally valid quantitative data support, significantly improving the accuracy of installation quality detection.

[0038] like Figure 2 As shown in Figure 3, the insulating protective cover of this invention undergoes deep functional modification at the physical structure level. The main body 1 of the insulating protective cover is made of modified polymer with high electrical insulation strength and weather resistance, ensuring long-term service capability under ultraviolet aging and acid rain corrosion environments. Furthermore, the opening edge of the main body 1 of the insulating protective cover integrates a magnetic self-locking component 2. This component is composed of several pairs of high-performance neodymium iron boron permanent magnets, whose magnetic pole arrangement follows the physical logic of opposite poles attracting each other. During the specific installation and covering action, when the two half-shells of the protective cover approach each other to the magnetic field induction critical point, the instantaneous attraction force generated by the magnetic self-locking component 2 will force the shell edges to align and achieve physical self-locking, eliminating the risk of insecure fastening due to uneven force from the operator. This self-locking logic of the hardware structure and the gap determination in the software algorithm form a perfect physical-digital mapping.

[0039] The insulating protective cover has a thermochromic coating 3 on its shell surface. This coating uses an organic color-changing microcapsule material with thermosensitive properties, exhibiting a clear color-changing warning characteristic between 60 and 80 degrees Celsius. In its specific physical mechanism, when the transformer terminals or circuit breaker fittings inside the protective cover experience excessive contact resistance or overload, causing a local temperature rise exceeding a preset threshold (e.g., reaching 70 degrees Celsius), the molecular arrangement structure within the thermochromic coating 3 undergoes a reversible change, inducing a physical shift in the reflectance spectrum. This results in a physical color change characterization from an initial color (e.g., green or blue) to a warning color (e.g., red or orange). This characterization provides a deterministic physical layer triggering logic for subsequent image color recognition models, achieving non-contact visualization of overheating conditions.

[0040] To address the issue that thermochromic signals are easily confused by ambient light in outdoor environments with varying color temperatures, the color recognition model module of this invention executes a precise automatic capture program. Furthermore, this module integrates the color-changing warning characteristics of the protective shield at 60-80 degrees Celsius, constructing a dedicated image color recognition model. During execution, the color recognition model module first performs color correction in the HSV (Hue, Saturation, Luminance) color space. This process extracts the chromaticity components of a preset reference object in the image (such as a specific area of ​​the hardware mounting bracket or a color correction block built into the protective shield), calculates the color cast of the current ambient light relative to a standard D65 light source, and performs inverse compensation. This eliminates the false redshift characteristics generated under low color temperature light in the morning or evening.

[0041] The color recognition model module extracts real-time color feature components from the surface of the insulating protective cover, specifically extracting the numerical distribution of the H channel. By calculating the offset of this component relative to a preset color threshold, the module can automatically capture and quantify color change signals caused by temperature anomalies. This quantitative analysis model transforms physical-level reflectance spectral changes into digital fault semantics, realizing an integrated logical closed loop of installation accuracy monitoring and operational fault early warning. Furthermore, the color recognition model module also integrates time series analysis logic. By comparing color vectors within multiple consecutive sampling periods, it eliminates instantaneous color abrupt changes caused by birds flying by or light spot flickering, ensuring the absolute stability of alarm commands.

[0042] The edge-to-edge collaborative interaction module described in this invention is responsible for constructing the computational and communication links of the entire monitoring system. In the scenario of distributed power distribution network inspection, uploading all high-resolution images to the cloud in real time would face enormous communication bandwidth pressure and processing latency. To address this, the edge-to-edge collaborative interaction module executes deep collaborative logic. In terms of computational logic allocation, the inspection terminal (such as a smart handheld device or an onboard unit of a drone) acts as the edge side to perform preliminary analysis of a lightweight model. This lightweight model is obtained by performing parameter pruning and knowledge distillation techniques on a deep neural network in the cloud. While retaining the core feature extraction capability, its computational load is compressed to less than 10% of the original model.

[0043] After capturing an image, the edge device immediately performs rapid prediction. When the image analysis algorithm module or color recognition model module detects anomalies such as suspected installation gaps or suspected color shifts, the edge-edge collaborative interaction module immediately encapsulates a feature packet containing details of the anomaly and transmits it back to the central management platform via a high-priority wireless link (such as a dedicated 5G network slice). Leveraging its abundant computing resources, the central management platform calls upon the full deep learning model to perform deep verification and analysis of the anomaly feature packet. This architecture not only reduces the continuous dependence on mobile communication bandwidth but also achieves edge-edge collaborative response for image acquisition, analysis, and early warning, ensuring millisecond-level latency from detecting potential hazards to triggering alarms.

[0044] The protective cover image sample library described in this invention is the core foundation for ensuring the generalization ability and robustness of the model. Through long-term accumulation of operation and maintenance data, this sample library has established a set of protective cover status images covering different hardware topologies (such as tension clamps, equipment pile heads, and surge arrester bases), different terrain backgrounds (such as mountain forest edges, urban high-rise buildings, and deserts), and different meteorological conditions. Furthermore, the sample library fully utilizes data augmentation technology and generative adversarial networks (GANs) to simulate visual occlusion, spatial topological interference, and imaging degradation effects at different observation distances from different perspectives through physical simulation. This ensures that the monitoring model can accurately extract the common topological features of the protective covers when facing heterogeneous hardware scenarios. Through continuous online evolution of the sample library, the system achieves seamless integration with the power grid intelligent operation and maintenance system, possessing extremely strong scene adaptability.

[0045] To more clearly demonstrate the technical advantages of this invention, the following is a data-driven description of the actual operation of the system through a complete embodiment.

[0046] 【Example】

[0047] In a 10kV urban power distribution network renovation project, the image analysis-based insulation cover installation monitoring system described in this invention was deployed. This implementation involved 500 sets of insulation covers equipped with magnetic self-locking components 2 and thermochromic coatings 3.

[0048] Installation monitoring phase: Inspectors collect images after installation using handheld terminals. The image preprocessing module completes adaptive histogram equalization and deblurring within 0.2 seconds. The image analysis algorithm module performs topological feature scanning on the opening location. The system automatically detects three installation defects, with pixel-level gap determination showing a gap width of 3.5mm (far exceeding the 0.2mm threshold in the closed state), and the edge features appearing broken. The system immediately displays an alarm on the handheld terminal screen via the edge-to-edge collaborative interaction module, indicating that the magnetic attraction is not engaged. The installer immediately resets and corrects the issue. After correction, the similarity comparison passes, and the installation is deemed complete.

[0049] Operational early warning phase: During the summer peak electricity consumption period three months after operation, the system routinely scanned the low-voltage side protective cover of a transformer using the color recognition model module. The color correction logic identified the ambient light offset, eliminating the influence of sunset red light. The measured H component value shifted from the initial 120 (green) to 15 (red area), exceeding the predicted threshold of 70%.

[0050] Response loop: The edge-to-edge collaborative interaction module uploaded the red anomaly packet to the cloud center. The cloud model confirmed that the discoloration was caused by temperature rise, determining that the internal hardware temperature was approximately 75 degrees Celsius. The center's intelligent operation and maintenance system issued an emergency repair command. On-site measurement revealed that the actual temperature of the hardware was 76.2 degrees Celsius due to increased contact resistance caused by loose bolts, successfully preventing a burnt-out lead wire failure.

[0051] To demonstrate the non-obviousness of this invention, the following comparative examples are constructed. Through analysis of the comparative data tables, the technological advancements of this invention are quantitatively presented.

[0052] [Comparative Example]

[0053] The comparison uses the current mainstream method of manual inspection of power distribution networks combined with traditional fixed threshold image recognition. This method lacks the semantic perception preprocessing logic of this invention and does not have magnetic self-locking hardware and active color-changing recognition algorithm.

[0054] Installation verification: Relying on visual inspection by inspection personnel using binoculars is highly susceptible to the influence of the shooting angle. The image algorithm uses simple edge detection, which often misinterprets shadows as installation gaps under shadow interference.

[0055] Status monitoring: Manual temperature measurement via infrared thermal imager is limited by the inspection cycle and cannot achieve all-weather color change warning. When recognizing colors in the image, color correction is not performed, and false color temperature alarms are frequently generated at dusk.

[0056] Table 1: Comparison of data between embodiments of the present invention and comparative examples

[0057]

[0058] The comparative data clearly demonstrates that the technical solution described in this invention achieves a qualitative leap in monitoring efficiency through the reconstruction of software algorithms and improvements in hardware structure. Particularly in terms of installation identification accuracy and detection precision, this invention reduces the false alarm rate by an order of magnitude through pixel-level topology scanning. Furthermore, in terms of fault early warning, this invention transforms the previously periodic, passive sampling inspection into an active early warning system based on physical color changes, significantly reducing the risks to power grid operation.

[0059] Furthermore, the image preprocessing module of this invention employs a special kernel estimation logic when handling motion blur. During the movement of the computation node, the parameters of the deconvolution kernel are adjusted in real time based on feedback from the built-in inertial sensor. This detail ensures that even during rapid drone flight inspections, the acquired images of the insulating protective cover still meet the requirements for pixel-level topology analysis. In addition, during color feature extraction, the color recognition model module not only extracts the single H component but also combines the saturation changes of the S channel to identify the subtle evolution trend of the color-changing coating under different temperature gradients, thereby achieving multi-level early warning.

[0060] The insulation protective cover installation monitoring system described in this invention exhibits extremely high synergy in its physical connections and logical transmission. The semantic awareness extraction logic ensures that the hardware can understand the spatial topological relationship between the fittings and the protective cover. For example, when analyzing the installation status at the transformer terminal, the image analysis algorithm module can automatically identify the cylindrical features of the terminal and, based on this, establish a polar coordinate system to perform a continuous scan of the open radius pi. This specific algorithmic reconstruction ensures the system's absolute awareness of key states such as proper installation, secure self-locking, and color-changing warnings.

[0061] In the deep application of edge-end collaboration, the edge-end collaboration interaction module adopts a dynamic weight update protocol. The central management platform re-executes knowledge distillation training based on the very few newly generated misjudgment cases in the protective cover image sample library, and pushes the updated lightweight model incrementally to all inspection terminals. This self-evolution capability ensures that the system can maintain a very high recognition rate as the power grid equipment models are updated and iterated.

[0062] Furthermore, in the manufacturing process of the insulating protective cover described in this invention, the magnetic self-locking component 2 is sealed within a special flame-retardant silicone layer to prevent magnetic attenuation due to electrochemical corrosion. The thermochromic coating 3 adopts a multi-layer encapsulation structure, with a highly reflective white reflective layer at the bottom, a thermosensitive color-changing layer in the middle, and an ultraviolet-resistant transparent fluorocarbon protective layer at the top. This three-layer physical structure corresponds to the color correction logic in the color recognition model, ensuring that the color recognition signal still has an extremely high signal-to-noise ratio even under strong direct sunlight.

[0063] This invention transforms complex power grid inspection operations into a computable digital model by atomically decomposing the installation and monitoring logic. The image acquisition module captures photon signals, the image preprocessing module enhances signal purity, the image analysis algorithm module performs geometric topology determination, the color recognition model module performs spectral characterization analysis, and the edge-end collaborative interaction module performs computational load allocation. Each step directly addresses the technical deficiencies mentioned in the background section. Through decoupling and reconstruction at the mechanistic level, intelligent control of the entire power grid equipment protection process is ultimately achieved.

[0064] In summary, this invention constructs a monitoring scheme for distribution network protection devices based on deep fusion of image semantics and physical coupling. Its core lies in enhancing environmental robustness through optimized image preprocessing technology, solving the challenge of refined identification using specific image analysis algorithms, achieving automatic temperature rise capture using a color recognition model, and realizing large-scale application of the system through lightweight end-edge collaboration and a generalized sample library. This invention not only resolves numerous contradictions in the background technology from a technical perspective but also provides a highly legally certain and technologically advanced implementation path for the operation and maintenance of smart grid equipment. The full lifecycle monitoring data generated by the system further provides a refined underlying data stream for the full lifecycle management and reliability evaluation of power grid assets.

[0065] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the present invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. An image analysis based insulation boot installation monitoring system comprising an insulation boot, characterized in that, The system also includes: The image acquisition module is used to acquire raw image signals containing the physical shape of the insulating protective cover, and convert the raw image signals into a digital image signal stream to be processed; An image preprocessing module is used to perform adaptive illumination compensation and multi-scale edge-preserving filtering on the digital image signal stream to eliminate ambient light and shadow interference and enhance the geometric contour features of the insulating protective cover, generating a high-fidelity feature image. The image analysis algorithm module is used to extract the topological features corresponding to the opening of the insulating protective cover in the high-fidelity feature image, determine the physical closing state of the magnetic self-locking component (2) of the insulating protective cover by identifying the pixel gap width, and output the installation position determination result; The logical process by which the image analysis algorithm module extracts topological features and determines the physical closure state includes: Extract the curvature, gap width, and shadow distribution pattern of the opening edge of the insulating protective cover in the pixel coordinate system; The boundary line between the two half-shells of the insulating protective cover is located by an edge detection operator, and the Euclidean distance between the edge pixels on both sides is calculated. When the Euclidean distance meets the preset closure threshold and the edge features present a continuous closed trajectory in the topological space, the physical closure state is determined to be a self-locking state. When discrete shadow transition points or non-continuous pixel holes exist in the edge features, the physical closure state is determined to be an incomplete installation state, and feature vector comparison is performed through the weighted cross-entropy loss function to output the determination result. The color recognition model module is used to obtain the color feature components of the surface of the insulating protective cover, and in the HSV color space, according to the offset of the color feature components relative to the preset color threshold, identify the color change characterization of the thermochromic coating (3) of the insulating protective cover, and output the temperature rise warning signal during operation. The edge-to-edge collaborative interaction module is used to perform preliminary identification of the topological features or color-changing representations on the edge side using a lightweight algorithm, and when an abnormal state is detected, it encapsulates an abnormal local detail feature package and transmits the abnormal local detail feature package back to the central management platform through a communication link with a preset priority.

2. The image analysis-based monitoring system for the installation of an insulating protective cover according to claim 1, characterized in that, The logical process by which the image preprocessing module performs adaptive illumination compensation includes: The digital image signal stream is divided into several non-overlapping sub-region blocks; For each sub-region block, its cumulative distribution function is calculated independently based on the pixel grayscale distribution characteristics. By utilizing a contrast-limited mechanism to perform mapping processing on the cumulative distribution function, the amplification of background noise is suppressed and the pixel brightness distribution is remapped, thereby achieving balanced extraction of features from different exposure areas.

3. The image analysis-based monitoring system for the installation of an insulating protective cover according to claim 1, characterized in that, The logical process by which the image preprocessing module performs multi-scale edge-preserving filtering includes: A spatial linear model of the guiding image is established, and the original image signal is used as guiding information to perform a linear transformation on the digital image signal stream; In a multi-scale architecture, high-frequency random noise is identified and filtered out by setting different receptive field ranges, while retaining gradient details that reflect the mechanical edges of the insulating shield, thereby enhancing the boundary clarity of the geometric contour features.

4. The image analysis-based monitoring system for the installation of an insulating protective cover according to claim 1, characterized in that, The image preprocessing module also includes a motion deblurring unit and a dehazing unit: The motion deblurring unit uses the spatially variable point diffusion function to model the image trailing caused by relative displacement during the acquisition process, and uses the deconvolution algorithm to converge the diffusion energy to the original pixel site. The dehazing unit uses dark channel prior logic to estimate the global atmospheric light intensity and transmittance map, and recovers the contrast and saturation of the high-fidelity feature image from the damaged image. The image preprocessing module also utilizes color space segmentation logic to logically remove vegetation background noise from the digital image signal stream by identifying the threshold of the green component corresponding to the environmental vegetation in the color space.

5. The image analysis-based monitoring system for the installation of an insulating protective cover according to claim 1, characterized in that, The logical process by which the color recognition model module identifies color change representations includes: Extract the chromaticity components of a preset reference object from the image, calculate the color cast of the current ambient light relative to the standard light source, and perform inverse color compensation to correct the ambient color temperature shift. The H channel numerical distribution of the surface of the insulating protective cover is extracted as the color feature component in the color-corrected image. The offset of the color feature components relative to the preset color change threshold is calculated, and the color vectors in multiple consecutive sampling periods are compared in combination with time series analysis logic. After excluding instantaneous color changes, the temperature anomaly signal is quantified, so as to realize the integrated logical association between the installation accuracy judgment result and the operating temperature rise warning signal.

6. The image analysis-based monitoring system for the installation of an insulating protective cover according to claim 1, characterized in that, The process by which the edge-end collaborative interaction module executes edge-end collaborative logic includes: A lightweight model, processed with parametric pruning and knowledge distillation, is run on the edge side; The lightweight model performs preliminary feature extraction and state prediction on the digital image signal stream. When the recognition result hits the preset abnormal feature library, it triggers the encapsulation action of the abnormal local detail feature package. The abnormal local detail features include the topological anomaly region image at the opening, the color vector data of the color anomaly region, and the preliminary identification results of the edge side. After receiving the abnormal local detail feature package, the central management platform calls the full deep learning model to perform depth verification.

7. The image analysis-based monitoring system for the installation of an insulating protective cover according to claim 1, characterized in that, The system also includes a protective shield image sample library, which is used to optimize the system's recognition generalization ability. The logical process of this library includes: The storage contains a set of images of the protective shield status under different hardware topologies, different terrain backgrounds, and different weather conditions; Generative adversarial networks are used to simulate imaging degradation effects under different viewpoint occlusion, spatial topological interference, and different observation distances to generate enhanced sample data. The image analysis algorithm module and the color recognition model module are iteratively trained based on the enhanced sample data to establish a common topology extraction logic for heterogeneous hardware scenarios.

8. The image analysis-based monitoring system for the installation of an insulating protective cover according to claim 1, characterized in that, The workflow for the system to perform closed-loop monitoring throughout its entire lifecycle includes: During the installation phase, the image analysis algorithm module performs topological scanning and pixel-level gap determination on the position of the magnetic self-locking component (2), and records the initial installation state data after the determination is passed; During the operation phase, it automatically switches to the operation monitoring mode, where the color recognition model module continuously monitors the color vector of the thermochromic coating (3), and outputs an alarm through the edge-to-edge collaborative interaction module when the color offset exceeds the warning threshold.

9. An image-analysis-based insulating protective cover, applicable to the image-analysis-based insulating protective cover installation monitoring system as described in any one of claims 1-8, characterized in that, The insulating protective cover includes: The main body of the protective cover (1) is made of a modified polymer with insulating strength and weather resistance. The magnetic self-locking component (2) is integrated into the opening edge of the protective cover body (1). The magnetic self-locking component (2) is composed of several pairs of high-performance neodymium iron boron permanent magnet materials. The magnetic pole arrangement of the neodymium iron boron permanent magnet materials follows the logic of opposite poles attracting each other, and is used to generate an automatic attraction force to achieve mechanical locking when the opening edge approaches the magnetic field induction critical point. A thermochromic coating (3) is applied to the surface of the protective cover body (1). The thermochromic coating (3) uses a microcapsule material with thermosensitive color development characteristics. The microcapsule material is used to generate a reflectance spectral shift when the internal temperature rises above a preset threshold, thereby achieving a physical color change from the initial color to the warning color.

Images