Power inspection method and system based on intelligent image analysis of unmanned aerial vehicle

Abstract

The invention discloses an electric power inspection method and system based on intelligent image analysis of an unmanned aerial vehicle, and relates to the technical field of electric power inspection, and the method comprises the steps: collecting image data of electric power equipment, and carrying out the preprocessing; based on a depth feature detection algorithm, edge features and structural features of the power equipment in the image data are extracted, a target area and type are identified, and a position is marked; carrying out the pixel-level segmentation of the surface features of the power equipment on the target region of the power equipment through employing an anomaly detection algorithm, extracting a defect region of the power equipment, and evaluating the defect degree; and performing feature matching on the inspection image data in different time periods, obtaining state change features of the power equipment, generating a health state change trend, and generating an inspection report. Through the combination of the unmanned aerial vehicle and the deep learning algorithm, the inspection efficiency and intelligent level of the power equipment are greatly improved, the limitation of manual inspection and fixed monitoring means is solved, and powerful technical support is provided for intelligent maintenance of the power equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of electric power inspection, and in particular to an electric power inspection method and system based on unmanned aerial vehicle intelligent image analysis. Background Art

[0002] With the widespread application of power equipment, the safety and reliability of its operating status directly affect the stability and economy of the power system. However, traditional power inspection methods usually rely on manual inspection or fixed monitoring equipment, which have the following main shortcomings: manual inspection is inefficient, time-consuming, costly and difficult to cover equipment in complex terrain; fixed monitoring has poor flexibility, and fixed monitoring equipment is limited by the installation position and angle, making it difficult to fully monitor the operating status of the equipment, and the monitoring range is limited; image analysis methods are insufficient, and traditional image analysis mainly relies on rule-based detection methods, which cannot effectively deal with the diversity and miniaturization characteristics of equipment defects in complex scenarios.

[0003] In recent years, with the development of drone technology and computer vision technology, drone-based power inspection methods have gradually emerged. With its flexible flight capabilities, drones can efficiently obtain high-resolution image data of a wide range of power equipment. At the same time, combined with intelligent image analysis technology, it can automatically identify the type of equipment in the image, locate the location of the equipment, and detect defects. However, the current technology still has the following problems: it is unable to effectively remove redundant information from image data, resulting in increased data storage and processing costs; it lacks accurate identification and evaluation methods for power equipment defects, making it difficult to accurately determine the severity of defects; it lacks the ability to analyze state change trends based on multiple inspection data, and cannot provide long-term health status assessment for equipment maintenance decisions. Summary of the invention

[0004] Based on the above-mentioned shortcomings of the prior art, the purpose of the present invention is to provide a power inspection method and system based on drone intelligent image analysis to solve the above-mentioned technical problems.

[0005] To achieve the above object, the present invention provides the following technical solution: a power inspection method based on UAV intelligent image analysis, comprising:

[0006] S1: collecting image data of power equipment by using a drone equipped with a high-resolution camera, and preprocessing the image data;

[0007] S2: Based on the deep feature detection algorithm, edge features and structural features of the power equipment in the preprocessed image data are extracted, the target area of ​​the power equipment and the type of the power equipment are identified according to the edge features and the structural features, and the position of the power equipment is marked;

[0008] S3: using an anomaly detection algorithm to perform pixel-level segmentation of surface features of the power equipment on the target area of ​​the power equipment, extracting defective areas of the power equipment, and evaluating the degree of defects;

[0009] S4: Perform feature matching on inspection image data in different time periods to obtain the state change characteristics of power equipment and generate health state change trends;

[0010] S5: Generate inspection reports based on power equipment type, power equipment location, defect degree and health status change trend.

[0011] The present invention is further configured such that the preprocessing includes denoising, color correction and geometric correction.

[0012] The present invention is further configured that step S2 specifically includes:

[0013] The edge features and structural features of the power equipment are extracted from the input image data through a deep feature detection algorithm, wherein the edge features E are extracted through the shallow convolution layer of the deep convolution network, and the structural features S are extracted through the deep convolution layer;

[0014] Optimizing the edge feature E based on the gradient direction to obtain an enhanced edge feature E′;

[0015] According to the enhanced edge feature E′ and the structural feature S, the global feature G of the power equipment is generated through the feature joint optimization formula;

[0016] The closed area formed by the edge feature E is set as the target area of ​​the power equipment. The type of power equipment is identified through the classification function according to the global feature G, and the location of the power equipment is marked.

[0017] The present invention is further configured to optimize the edge feature E based on the gradient direction, and the optimization logic is: Among them, α is the enhancement coefficient, which is used to control the degree of enhancement of edge features by gradient optimization. is the co-variation term of image gradient and edge features;

[0018] The feature joint optimization formula is: G = γ·E′ + ∫ Ω β(x,y)·Sdxdy, where γ is the weight coefficient of the edge feature, Ω is the spatial area of ​​the image data, and β(x,y) is the spatial distribution weight of the structural feature S. is the gradient of the structural feature S at the coordinate (x, y);

[0019] The logic of power equipment type identification is: Among them, t * For the type of electrical equipment identified, is the set of power equipment types, ψ(·) is the classification function, θ cls is the classification function parameter.

[0020] The present invention is further configured that step S3 specifically includes:

[0021] The pixel values ​​of the target area of ​​the power equipment are set as pixel features, and a target feature matrix is ​​constructed;

[0022] By analyzing the pixel distribution of the target feature matrix, an adaptive feature matrix is ​​constructed;

[0023] Calculate the abnormality score of each pixel according to the target feature matrix and the adaptive feature matrix, and perform region segmentation to extract defective regions according to the abnormality score;

[0024] The defect degree of the defect area is evaluated according to the abnormality score.

[0025] The present invention is further configured to set the pixel values ​​of the target area of ​​the power equipment as pixel features, construct a target feature matrix, F surface (x,y)=I target (x,y), where F surface (x,y) is the target feature matrix, I target (x,y) is the pixel value of the image at coordinate (x,y);

[0026] The construction logic of the adaptive feature matrix is: Among them, P adaptive (x, y) is the adaptive feature matrix, K is the number of clustering categories, μ k is the feature center of the kth class, π k (x,y) is the weight distribution of the kth class, defined as

[0027] The calculation logic of the anomaly score for each pixel is: A(x,y) is the anomaly score of the pixel at coordinate (x,y), is the local gradient at coordinate (x, y);

[0028] According to the anomaly score, the defect area is extracted by regional segmentation, and the segmentation logic is: M(x,y) is the defect area segmentation value, T anomaly is the threshold of the anomaly score, defect area D = {(x, y) | M (x, y) = 1};

[0029] The calculation logic of defect degree assessment is: Among them, S severity Assess the score for the degree of defect.

[0030] The present invention is further configured that step S4 specifically includes:

[0031] Feature extraction is performed on the image data of the same power equipment in different time periods to generate a global feature sequence G of multiple time periods. t ;

[0032] For the global feature sequence G t Through the transformation matrix T t Align the global features of each time period to a unified reference coordinate system;

[0033] In the aligned global feature sequence, global feature matching across time is performed to calculate the state change features of the power equipment;

[0034] Arrange the state change characteristics of different time periods in chronological order to generate a state change characteristic sequence of power equipment;

[0035] The health status change trend matrix of the power equipment is calculated based on the status change characteristic sequence.

[0036] The present invention is further configured such that the logic of global feature alignment is is the global feature after alignment, T t =argmin T ∑ (x,y) ∥G t (x,y)-G 0 (x,y)∥ 2 , where G 0 (x,y) is the feature matrix at the reference time point;

[0037] The calculation logic of the state change feature is: Among them, ΔG t (x,y) is the state change feature, is the gradient of the state change feature at time t;

[0038] The characteristic sequence of power equipment state change S(x,y)={ΔG 1 (x,y),ΔG 2 (x,y),…,ΔG T (x, y)}. The present invention is further configured such that the calculation logic of the health status change trend matrix is: Among them, H t (x, y) is the health status change trend matrix, δ is the time decay factor;

[0039] For the trend matrix H t(x, y) is normalized to obtain the health status change coefficient, Among them, S health is the health status variation coefficient, and Ω is the pixel set of the power equipment area.

[0040] The present invention also provides a power inspection system based on unmanned aerial vehicle intelligent image analysis, the system comprising:

[0041] Image acquisition module: collects image data of power equipment through a drone equipped with a high-resolution camera and pre-processes the image data;

[0042] Equipment identification module: Based on the deep feature detection algorithm, the edge features and structural features of the power equipment in the preprocessed image data are extracted, the target area of ​​the power equipment and the type of the power equipment are identified according to the edge features and the structural features, and the position of the power equipment is marked;

[0043] Defect assessment module: using an anomaly detection algorithm to perform pixel-level segmentation of surface features of the power equipment on the target area of ​​the power equipment, extracting defect areas of the power equipment, and assessing the degree of defects;

[0044] Trend generation module: performs feature matching on inspection image data in different time periods, obtains the state change characteristics of power equipment, and generates health status change trends;

[0045] Report generation module: Generate inspection reports based on power equipment type, power equipment location, defect degree and health status change trend.

[0046] The present invention provides a power inspection method and system based on intelligent image analysis of unmanned aerial vehicles. The method collects image data of power equipment by using a high-resolution camera equipped on an unmanned aerial vehicle, and pre-processes the image data; extracts edge features and structural features of the power equipment in the pre-processed image data based on a deep feature detection algorithm, identifies a target area of ​​the power equipment and a type of the power equipment according to the edge features and the structural features, and marks the position of the power equipment; performs pixel-level segmentation of surface features of the power equipment on the target area of ​​the power equipment using an anomaly detection algorithm, extracts defective areas of the power equipment, and evaluates the degree of defects; performs feature matching on inspection image data of different time periods, obtains state change features of the power equipment, and generates a health state change trend; generates an inspection report according to the type of power equipment, the location of the power equipment, the degree of defects, and the health state change trend, and the beneficial effects produced include:

[0047] 1. High-precision defect identification and severity assessment: Through pixel-level segmentation technology and adaptive feature matrix construction algorithm, it can perform fine-grained analysis on the surface defect area of ​​power equipment, effectively detect common defects such as cracks and rust, and scientifically evaluate the degree of defects through calculation of abnormal scores and regional segmentation methods. Combined with adaptive distribution analysis technology, it improves the detection accuracy and robustness of various types of equipment surface defects;

[0048] 2. Support dynamic monitoring of equipment health status: Generate the characteristic sequence of health status change of power equipment through feature matching algorithm, and calculate the health status change trend matrix based on multiple inspection data, which can accurately evaluate the status change of power equipment. Introducing time series analysis can effectively predict the operating status of equipment and provide a quantitative basis for equipment maintenance and repair decisions.

[0049] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0051] Figure 1 A flowchart of a power inspection method based on intelligent image analysis of unmanned aerial vehicles is shown as an exemplary embodiment of the present invention;

[0052] Figure 2 The present invention is a schematic structural diagram of an electric power inspection system based on intelligent image analysis of unmanned aerial vehicles, showing an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0053] The following will describe the embodiments of the present invention with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention, not for limiting the scope of protection of the present invention.

[0054] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and thus the drawings only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0055] In the following description, numerous details are discussed to provide a more thorough explanation of the embodiments of the present invention. However, it is obvious to those skilled in the art that the embodiments of the present invention can be implemented without these specific details. In other embodiments, well-known structures and devices are shown in the form of block diagrams rather than in detail to avoid making the embodiments of the present invention difficult to understand.

[0056] Embodiment 1

[0057] A power inspection method based on UAV intelligent image analysis, such as Figure 1 As shown, including:

[0058] S1: collecting image data of power equipment by using a drone equipped with a high-resolution camera, and preprocessing the image data;

[0059] S2: Based on the deep feature detection algorithm, edge features and structural features of the power equipment in the preprocessed image data are extracted, the target area of ​​the power equipment and the type of the power equipment are identified according to the edge features and the structural features, and the position of the power equipment is marked;

[0060] S3: using an anomaly detection algorithm to perform pixel-level segmentation of surface features of the power equipment on the target area of ​​the power equipment, extracting defective areas of the power equipment, and evaluating the degree of defects;

[0061] S4: Perform feature matching on inspection image data in different time periods to obtain the state change characteristics of power equipment and generate health state change trends;

[0062] S5: Generate inspection reports based on power equipment type, power equipment location, defect degree and health status change trend.

[0063] Specifically, the camera equipment carried by the drone can collect images of power equipment in a large range, with a resolution of up to 4K or higher, ensuring the complete preservation of image details, facilitating subsequent precise analysis, and the drone transmits image data in real time to a ground workstation or cloud for storage and analysis through wireless communication, including 5G, Wi-Fi or a dedicated communication protocol. The present invention is further configured that the preprocessing includes denoising, color correction and geometric correction. Denoising: Non-local mean filtering, smoothing the random noise in the image based on pixel similarity, retaining the edge details of the image. Wavelet transform denoising, converting the image to the frequency domain, suppressing high-frequency noise, while retaining key features. Color correction: Eliminating color distortion caused by different lighting conditions (such as shadows, light intensity) and enhancing image contrast. Use grayscale world hypothesis algorithm or gamma correction and other methods to restore the true color distribution of the image. Geometric correction: Repairing geometric distortion caused by changes in camera angles through projection transformation, distortion correction and other methods to ensure the true proportion of the target device in the image. It is a prior art and will not be described here. Through the above design, high-quality image data can be efficiently acquired in different environments, providing solid technical support for the intelligent inspection and analysis of power equipment.

[0064] The present invention is further configured that step S2 specifically includes:

[0065] The edge features and structural features of the power equipment are extracted from the input image data through a deep feature detection algorithm, wherein the edge features E are extracted through the shallow convolution layer of the deep convolution network, and the structural features S are extracted through the deep convolution layer; specifically, the edge features and structural features of the image are extracted through the different hierarchical features of the deep convolutional neural network (CNN), respectively, in order to achieve a comprehensive description and accurate detection of the power equipment. The shallow convolution layer focuses on low-level visual features, including edges, corners and textures, which are particularly critical for capturing the contours or boundaries of the equipment. The deep convolution layer focuses on high-level semantic information, including complex shapes, patterns or textures, which are used to describe the internal structure or surface characteristics of the power equipment. The convolution kernel slides on the image, and specific low-level features are extracted by calculating the dot product between the convolution kernel and the local area of ​​the image. The subsequent layers of the deep convolution network learn more complex semantic information through more convolution operations based on the shallow features. The edge features extracted through the shallow convolution layer can effectively capture the geometric contours of the power equipment, providing a basis for the target positioning and detection of the equipment. By extracting structural features through deep convolutional layers, the complex texture and semantic information of the device surface can be accurately described, providing support for subsequent defect detection and classification.

[0066] The edge feature E is optimized based on the gradient direction to obtain an enhanced edge feature E′; the present invention is further configured to optimize the edge feature E based on the gradient direction, and the optimization logic is: Among them, α is the enhancement coefficient, which is used to control the degree of enhancement of edge features by gradient optimization. is the synergistic variation term of image gradient and edge feature; specifically, the edge feature E is enhanced by optimizing the gradient direction to generate a more significant edge feature E′. The weighted operation is combined with the strength of the edge feature itself to enhance the ability to describe the device boundary and improve the robustness of edge detection under complex backgrounds. The edge feature E represents the area where the pixel grayscale value in the input image changes significantly, usually corresponding to the boundary or contour of the object. The coordinated change term of the image gradient and the edge feature It represents the relationship between the gradient direction and the edge feature intensity, and reflects the synergy between the pixel intensity change and the boundary significance. The enhancement coefficient α is used to control the enhancement amplitude of the gradient optimization, and its value range is [0.1, 2.0]. By enhancing the edge features through gradient optimization, the boundary is made more significant in a complex background, thereby improving the accuracy of device detection.

[0067] According to the enhanced edge feature E′ and the structural feature S, the global feature G of the power equipment is generated by the feature joint optimization formula; the feature joint optimization formula is: G = γ·E′ + ∫ Ω β(x,y)·Sdxdy, where γ is the weight coefficient of the edge feature, Ω is the spatial area of ​​the image data, and β(x,y) is the spatial distribution weight of the structural feature S. is the gradient of the structural feature S at the coordinate (x, y); specifically, the weight coefficient γ of the edge feature is used to reflect the importance of the edge feature in the global feature, and its value range is [0.5, 2.0]. Ω β(x,y)·Sdxdy performs weighted accumulation on the spatial distribution of the structural feature S. By regional weighting, important areas are highlighted, including areas with device defects or high texture intensity, and irrelevant areas are reduced. By fusing edge features and structural features, the global features can accurately describe the geometric contour of the device and reflect the surface texture characteristics of the device.

[0068] The closed area formed by the edge feature E is set as the target area of ​​the power equipment. The type of power equipment is identified through the classification function according to the global feature G, and the location of the power equipment is marked. The logic of power equipment type identification is: Among them, t * is the identified power equipment type, T is the set of power equipment types, ψ(·) is the classification function, θ cls is the classification function parameter. Specifically, the classification function ψ(·) is used to analyze and classify global features and output the probability distribution of each device type. The classification function parameter θ clsIt represents the internal parameters of the classifier, which is obtained through learning and training, minimizing the classification error, and learning the features with strong characterization ability for the device type. It is an existing technology and will not be described here. The device type is classified and identified through the global feature G and the classification function ψ(·). By maximizing the output of the classification function, the device type t that best meets the feature G is selected. * ,Finally complete the labeling of the equipment type, confirm the location of the equipment type through the location of the UAV, and record the spatial coordinates of the target area.

[0069] The present invention is further configured that step S3 specifically includes:

[0070] The pixel values ​​of the target area of ​​the power equipment are set as pixel features to construct a target feature matrix; the present invention is further configured to set the pixel values ​​of the target area of ​​the power equipment as pixel features to construct a target feature matrix, F surface (x,y)=I target (x,y), where F surface (x,y) is the target feature matrix, I target (x, y) is the pixel value of the image at the coordinate (x, y); specifically, by extracting the pixel values ​​of the target area of ​​the power equipment, the target feature matrix F is constructed. surface (x, y), used to fully characterize the texture, color and pixel features of the device surface. The target feature matrix is ​​the mapping of image pixel values ​​on the spatial coordinates (x, y);

[0071] By analyzing the pixel distribution of the target feature matrix, an adaptive feature matrix is ​​constructed; the construction logic of the adaptive feature matrix is: Among them, P adaptive (x, y) is the adaptive feature matrix, K is the number of clustering categories, μ k is the feature center of the kth class, π k (x,y) is the weight distribution of the kth class, defined as By transforming the target feature matrix F surface Analyze the pixel distribution of (x, y) and construct the adaptive feature matrix P adaptive (x, y). The adaptive feature matrix clusters the pixels in the target area and divides the feature values ​​of the target area into multiple categories. Each pixel is assigned a weight according to its similarity with the feature center of each category, thereby generating a feature matrix containing spatial weight information, ∥F surface (x,y)-μ k ∥ 2 is the Euclidean distance between the pixel value and the feature center, indicating the similarity between the pixel and the category, π k(x, y) represents the weight of a pixel belonging to a certain category. The sum of all category weights is 1. The weight of each pixel is calculated based on the distance between its eigenvalue and the K feature centers. The closer the distance, the greater the weight. The adaptive feature matrix combines the feature center of each category with the corresponding weight through weighted summation. The generated feature matrix can reflect the pixel distribution characteristics and category weight information of the target area.

[0072] The abnormal score of each pixel is calculated according to the target feature matrix and the adaptive feature matrix, and the defect area is extracted by performing region segmentation according to the abnormal score; the calculation logic of the abnormal score of each pixel is: A(x,y) is the anomaly score of the pixel at coordinate (x,y), is the local gradient at the coordinate (x, y); the defective area is extracted by performing regional segmentation according to the anomaly score, and the segmentation logic is: M(x,y) is the defect area segmentation value, T anomaly is the threshold of the abnormal score, the defect area D = {(x, y) | M (x, y) = 1}; by surface (x,y) and the adaptive feature matrix P adaptive (x, y) and calculate the anomaly score A(x, y) of each pixel. anomaly The abnormal area is determined by comparison, and the segmentation result D of the defect area is generated. The purpose is to separate the abnormal area on the surface of the equipment from the target area to provide support for defect detection and severity assessment; the abnormal score of each pixel is calculated by calculating the difference between the target feature matrix and the adaptive feature matrix, measuring the degree of deviation of the pixel value, and adding the local gradient Enhance the anomaly detection capability of boundary features; M(x,y) is the defect area segmentation value, specifically the binary mask of the abnormal area segmentation result, 1 represents the abnormal area, and 0 represents the normal area. Traverse the binary segmentation results, collect all abnormal pixels, and form a complete defect area D;

[0073] The defect degree of the defect area is evaluated according to the abnormal score, and the calculation logic of the defect degree evaluation is: Among them, S severity is the defect degree evaluation score; specifically, by comparing the anomaly score A(x,y) and the local gradient The weighted integral calculation of is used to evaluate the overall severity of the defect area. The key point is to combine the pixel-level anomaly score with the boundary significance to comprehensively reflect the area, distribution characteristics and gradient strength of the defect area, thereby quantifying the severity of the defect.

[0074] The present invention is further configured that step S4 specifically includes:

[0075] Feature extraction is performed on the image data of the same power equipment in different time periods to generate a global feature sequence G of multiple time periods. t ;

[0076] For the global feature sequence G t Through the transformation matrix T t Align the global features of each time period to a unified reference coordinate system; the present invention is further configured such that the logic of global feature alignment is is the global feature after alignment, T t =argmin T ∑ (x,y) ∥G t (x,y)-G 0 (x,y)∥ 2 , where G 0 (x, y) is the feature matrix of the reference time point; specifically, by extracting features from the image data collected by the power equipment in different time periods, a global feature sequence G for multiple time periods is generated. t Then, through the transformation matrix T t Align the global features of each time period to a unified reference coordinate system so that the data of different time periods are consistent. The transformation matrix T t The global feature matrix at time point t is compared with the feature matrix G at the reference time point 0 The optimal transformation matrix for alignment, ∥G t (x,y)-G 0 (x,y)∥ 2 is the Euclidean distance of the feature matrix, which is used to measure the similarity of two feature matrices. T To optimize the operation, we search for a transformation matrix that can minimize the difference in feature matrices. Through alignment operations, we eliminate the effects of displacement and rotation in the image data, so that global features at different time periods can be compared in a unified reference coordinate system.

[0077] In the aligned global feature sequence, global feature matching across time is performed to calculate the state change features of the power equipment; the calculation logic of the state change features is: Among them, ΔG t (x,y) is the state change feature, is the gradient of the state change feature at time t; specifically, the state change feature ΔG of the device is calculated by matching the feature matrices at different time points in the aligned global feature sequence t (x, y), the main purpose is to quantify the degree of state change of the device between consecutive time points, so as to describe the change of the health state of the device; Calculate the pixel difference between the aligned feature matrices at time points t and t-1 to quantify the state change of the device between consecutive time points. It is the gradient of the alignment feature matrix at time t, which is used to reflect the local change rate of pixels, weight the state change, reduce the interference of boundary noise with large feature gradient on the result, highlight the significant changes in smooth areas, and accurately quantify the degree of change of the device in continuous time periods through the difference of features between time points and gradient weighting;

[0078] The state change characteristics of different time periods are arranged in chronological order to generate a power equipment state change characteristic sequence; the power equipment state change characteristic sequence S(x,y)={ΔG 1 (x,y),ΔG 2 (x,y),…,ΔG T (x, y)}. By arranging the state change characteristics of different time periods in chronological order, a power equipment state change characteristic sequence S(x, y) is formed to describe the state evolution of the equipment in the time series and comprehensively describe the state change of the equipment in the time series;

[0079] The health state change trend matrix of the power equipment is calculated according to the state change characteristic sequence. The present invention is further configured such that the calculation logic of the health state change trend matrix is: Among them, H t (x, y) is the health status change trend matrix, δ is the time decay factor; for the trend matrix H t (x, y) is normalized to obtain the health status change coefficient, Among them, S health is the health state change coefficient, Ω is the pixel set of the power equipment area; specifically, the health state change trend matrix H of the power equipment is calculated by performing time-weighted accumulation on the state change feature sequence t (x, y). Through further normalization processing, the health status change coefficient S is generated health , which is used to quantify the overall health status change trend of the equipment, and uses the time decay factor δ to calculate the state change characteristic sequence ΔG i (x, y) weighted accumulation to generate the health status change trend matrix H t (x, y), the health status change trend matrix H t (x, y) is normalized and calculated to obtain the health status change coefficient S health The time decay factor δ is used to control the time weight distribution, and its value range is [0.01, 0.5]. The health status change trend matrix provides basic data for the long-term trend analysis of health status and future status prediction. The health status change coefficient S healthProvides a quantitative metric that is easy to compare and is used to quickly assess the overall health of the device.

[0080] Embodiment 2

[0081] See also Figure 2 , the exemplary power inspection system based on UAV intelligent image analysis includes:

[0082] Image acquisition module: collects image data of power equipment through a drone equipped with a high-resolution camera and pre-processes the image data;

[0083] Equipment identification module: Based on the deep feature detection algorithm, the edge features and structural features of the power equipment in the preprocessed image data are extracted, the target area of ​​the power equipment and the type of the power equipment are identified according to the edge features and the structural features, and the position of the power equipment is marked;

[0084] Defect assessment module: using an anomaly detection algorithm to perform pixel-level segmentation of surface features of the power equipment on the target area of ​​the power equipment, extracting defect areas of the power equipment, and assessing the degree of defects;

[0085] Trend generation module: performs feature matching on inspection image data in different time periods, obtains the state change characteristics of power equipment, and generates health status change trends;

[0086] Report generation module: Generate inspection reports based on power equipment type, power equipment location, defect degree and health status change trend.

[0087] It should be noted that the electric power inspection system based on drone intelligent image analysis provided in the above embodiment and the electric power inspection method based on drone intelligent image analysis provided in the above embodiment belong to the same concept, wherein the specific manner in which each module and unit performs the operation has been described in detail in the method embodiment and will not be repeated here. In actual application, the electric power inspection system based on drone intelligent image analysis provided in the above embodiment can distribute the above functions to different functional modules as needed, that is, divide the internal structure of the system into different functional modules to complete all or part of the functions described above, and this is not limited here.

[0088] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0089] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0090] In this application, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can be represented by: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple.

[0091] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0092] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0093] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0094] In the several embodiments provided in the present application, it should be understood that the disclosed system can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0095] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0096] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0097] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage media include: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks or optical disks.

[0098] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A power inspection method based on UAV intelligent image analysis, characterized in that: include: S1: collecting image data of power equipment by using a drone equipped with a high-resolution camera, and preprocessing the image data; S2: Based on the deep feature detection algorithm, edge features and structural features of the power equipment in the preprocessed image data are extracted, the target area of ​​the power equipment and the type of the power equipment are identified according to the edge features and the structural features, and the position of the power equipment is marked; S3: using an anomaly detection algorithm to perform pixel-level segmentation of surface features of the power equipment on the target area of ​​the power equipment, extracting defective areas of the power equipment, and evaluating the degree of defects; S4: Perform feature matching on inspection image data in different time periods to obtain the state change characteristics of power equipment and generate health state change trends; S5: Generate inspection reports based on power equipment type, power equipment location, defect degree and health status change trend.

2. The power inspection method based on UAV intelligent image analysis according to claim 1 is characterized in that: The preprocessing includes denoising, color correction and geometric correction.

3. The power inspection method based on UAV intelligent image analysis according to claim 1 is characterized in that: Step S2 specifically includes: The edge features and structural features of the power equipment are extracted from the input image data through a deep feature detection algorithm, wherein the edge features E are extracted through the shallow convolution layer of the deep convolution network, and the structural features S are extracted through the deep convolution layer; Optimizing the edge feature E based on the gradient direction to obtain an enhanced edge feature E′; According to the enhanced edge feature E′ and the structural feature S, the global feature G of the power equipment is generated through the feature joint optimization formula; The closed area formed by the edge feature E is set as the target area of ​​the power equipment. The type of power equipment is identified through the classification function according to the global feature G, and the location of the power equipment is marked.

4. The power inspection method based on UAV intelligent image analysis according to claim 3 is characterized in that: The edge feature E is optimized based on the gradient direction, and the optimization logic is: Among them, α is the enhancement coefficient, which is used to control the degree of enhancement of edge features by gradient optimization. It is the co-variation term of image gradient and edge features; The feature joint optimization formula is: G = γ·E′ + ∫ Ω β(x,y)·Sdxdy, where γ is the weight coefficient of the edge feature, Ω is the spatial area of ​​the image data, and β(x,y) is the spatial distribution weight of the structural feature S. is the gradient of the structural feature S at the coordinate (x, y); The logic of power equipment type identification is: Among them, t * For the type of electrical equipment identified, is the set of power equipment types, ψ(·) is the classification function, θ cls is the classification function parameter.

5. The power inspection method based on UAV intelligent image analysis according to claim 1 is characterized in that: Step S3 specifically includes: The pixel values ​​of the target area of ​​the power equipment are set as pixel features, and a target feature matrix is ​​constructed; By analyzing the pixel distribution of the target feature matrix, an adaptive feature matrix is ​​constructed; Calculate the abnormality score of each pixel according to the target feature matrix and the adaptive feature matrix, and perform region segmentation to extract defective regions according to the abnormality score; The defect degree of the defect area is evaluated according to the abnormality score.

6. The power inspection method based on UAV intelligent image analysis according to claim 5 is characterized in that: The pixel values ​​of the target area of ​​the power equipment are set as pixel features, and the target feature matrix is ​​constructed. surface (x,y)=I target (x,y), where F surface (x,y) is the target feature matrix, I target (x,y) is the pixel value of the image at coordinate (x,y); The construction logic of the adaptive feature matrix is: Among them, P adaptive (x, y) is the adaptive feature matrix, K is the number of clustering categories, μ k is the feature center of the kth class, π k (x,y) is the weight distribution of the kth class, defined as The calculation logic of the anomaly score for each pixel is: A(x,y) is the anomaly score of the pixel at coordinate (x,y), is the local gradient at coordinate (x, y); According to the anomaly score, the defect area is extracted by regional segmentation, and the segmentation logic is: M(x,y) is the defect area segmentation value, T anomaly is the threshold of the anomaly score, defect area D = {(x, y) | M (x, y) = 1}; The calculation logic of defect degree assessment is: Among them, S severity Assess the score for the degree of defect.

7. The power inspection method based on UAV intelligent image analysis according to claim 1 is characterized in that: Step S4 specifically includes: Feature extraction is performed on the image data of the same power equipment in different time periods to generate a global feature sequence G of multiple time periods. t ; For the global feature sequence G t Through the transformation matrix T t Align the global features of each time period to a unified reference coordinate system; In the aligned global feature sequence, global feature matching across time is performed to calculate the state change features of the power equipment; Arrange the state change characteristics of different time periods in chronological order to generate a state change characteristic sequence of power equipment; The health status change trend matrix of the power equipment is calculated based on the status change characteristic sequence.

8. The power inspection method based on UAV intelligent image analysis according to claim 7 is characterized in that: The logic of global feature alignment is is the global feature after alignment, T t =argmin T ∑ (x,y) ∥G t (x,y)-G0(x,y)∥ 2 , where G0(x,y) is the feature matrix at the reference time point; The calculation logic of the state change feature is: Among them, ΔG t (x,y) is the state change feature, is the gradient of the state change feature at time t; The characteristic sequence of power equipment state change S(x,y)={ΔG1(x,y),ΔG2(x,y),…,ΔG T (x,y)}.

9. The power inspection method based on UAV intelligent image analysis according to claim 8 is characterized in that: The calculation logic of the health status change trend matrix is: Among them, H t (x, y) is the health status change trend matrix, δ is the time decay factor; For the trend matrix H t (x, y) is normalized to obtain the health status change coefficient, Among them, S health is the health status variation coefficient, and Ω is the pixel set of the power equipment area.

10. A power inspection system based on drone intelligent image analysis, used to implement a power inspection method based on drone intelligent image analysis as described in any one of claims 1 to 9, characterized in that: include: Image acquisition module: collects image data of power equipment through a drone equipped with a high-resolution camera and pre-processes the image data; Equipment identification module: Based on the deep feature detection algorithm, the edge features and structural features of the power equipment in the preprocessed image data are extracted, the target area of ​​the power equipment and the type of the power equipment are identified according to the edge features and the structural features, and the position of the power equipment is marked; Defect assessment module: using an anomaly detection algorithm to perform pixel-level segmentation of surface features of the power equipment on the target area of ​​the power equipment, extracting defect areas of the power equipment, and assessing the degree of defects; Trend generation module: performs feature matching on inspection image data in different time periods, obtains the state change characteristics of power equipment, and generates health status change trends; Report generation module: Generate inspection reports based on power equipment type, power equipment location, defect degree and health status change trend.

Images