Unmanned aerial vehicle project acceptance method based on ViT detection algorithm

Abstract

The invention relates to an unmanned aerial vehicle project acceptance method based on a ViT detection algorithm, and the method comprises the steps: collecting the panoramic image data of a distribution network in real time through an image collection sensor, recognizing a plurality of key regions of the distribution network, and obtaining the data of a key acceptance region; adjusting flight path parameters of the unmanned aerial vehicle by considering the spatial positions and environmental obstacles of the plurality of key areas, optimizing the coverage range of image acquisition, and obtaining flight path adjustment parameters; performing semantic segmentation on the image acquired by the unmanned aerial vehicle, analyzing and identifying each part of the distribution network, and generating distribution network feature identification data; constructing a three-dimensional model of each part of the distribution network by calculating the geometrical characteristics of each part of the distribution network so as to generate a three-dimensional model of the distribution network; analyzing the structural integrity of the distribution network, evaluating the influence on the security of the distribution network, and generating a structural integrity analysis result; and evaluating the matching degree of the distribution network quality and the engineering specification, and generating an unmanned aerial vehicle engineering acceptance record. The method is beneficial for improving the acceptance efficiency, quality and safety of the distribution network project.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering acceptance, and in particular to a method for unmanned aerial vehicle engineering acceptance based on a ViT detection algorithm. Background Art

[0002] Engineering acceptance technology focuses on the evaluation and review process after the completion of the project. During the power project review process, engineering acceptance technology can ensure that construction activities meet the quality standards and regulatory requirements of the power industry. Through multiple steps such as structural integrity testing, safety assessment, environmental impact assessment, and compliance inspection, the project is ensured to be safe, functional, and legally compliant. Special attention is paid to combining load testing, thermal imaging, and acoustic wave testing to ensure that the structure and function of the power project meet the strict requirements of design and regulations.

[0003] Traditional project acceptance methods have significant limitations in the field of power engineering. They rely on manpower or specific tools and equipment, and are particularly limited in areas that are difficult to reach directly. This limitation makes it impossible for acceptance work to fully cover multiple key areas in the distribution network, making it difficult to collect standardized and accurate data, and unable to provide high-quality data for subsequent distribution network project acceptance analysis. Summary of the invention

[0004] The purpose of the present invention is to provide a UAV project acceptance method based on the ViT detection algorithm, which is conducive to improving the efficiency, quality and safety of distribution network project acceptance.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is: a UAV engineering acceptance method based on ViT detection algorithm, comprising the following steps:

[0006] The image acquisition sensor on the drone collects panoramic image data of the distribution network in real time, and analyzes the geometric and structural features of the image to identify multiple key areas of the distribution network and obtain key acceptance area data;

[0007] According to the key acceptance area data, considering the spatial positions and environmental obstacles of multiple key areas, adjusting the flight path parameters of the UAV, optimizing the coverage of image acquisition, and obtaining the flight path adjustment parameters;

[0008] Adjusting parameters according to the flight path, performing semantic segmentation on the images collected by the drone, analyzing and identifying various parts of the distribution network, and generating distribution network feature recognition data;

[0009] Based on the distribution network feature recognition data, by calculating the geometric features of each part of the distribution network, a three-dimensional model of each part of the distribution network is constructed, thereby generating a three-dimensional model of the distribution network;

[0010] Based on the three-dimensional model of the distribution network, the structural integrity of the distribution network is analyzed, and the impact on the safety of the distribution network is evaluated to generate a structural integrity analysis result;

[0011] According to the structural integrity analysis results, by comparing with the safety specifications and quality standards of distribution network acceptance, the matching degree between distribution network quality and engineering specifications is evaluated, and the UAV engineering acceptance record is generated.

[0012] Furthermore, the panoramic image data of the distribution network is collected in real time by the image acquisition sensor on the drone. By analyzing the geometric and structural features of the image, multiple key areas of the distribution network are identified and the key acceptance area data is obtained, specifically:

[0013] The image data of the distribution network area is collected by the image acquisition sensor on the drone, a panoramic image of the distribution network area is obtained, and panoramic image data is generated;

[0014] Based on the panoramic image data, analyzing geometric and structural features in the distribution network image, identifying structural boundaries and key construction elements, and generating key construction element data;

[0015] Based on the key construction element data, by evaluating the importance and structural complexity of multiple areas, multiple key acceptance areas are analyzed and identified to generate key acceptance area data.

[0016] Furthermore, based on the panoramic image data, the geometric and structural features in the distribution network image are analyzed to identify the structural boundaries and key building elements, including beams, columns and supporting structures, and generate key building element data, specifically:

[0017] Based on the panoramic image data, the image is preprocessed to enhance the geometric and structural features in the image;

[0018] Apply edge detection algorithms to identify the outline of the structure, use geometric pattern matching technology to identify multiple key building elements, set feature vectors for each key building element and perform database comparison;

[0019] The structure recognition scoring formula is used to evaluate the structure recognition results:

[0020] Among them, S z represents the structure recognition score, is the weight of the i0th key building element, is the matching degree of the key building element in the image, and n0 is the number of types of key building elements, which are used to locate and identify the key building elements in the image and generate key building element data;

[0021] According to the key construction element data, by evaluating the importance and structural complexity of multiple areas, multiple key acceptance areas are analyzed and identified, and key acceptance area data is generated, specifically:

[0022] According to the key construction element data, the importance of multiple areas is ranked, and each area is evaluated using a complexity scoring formula considering structural complexity and safety factors;

[0023]

[0024] Among them, C z is the complexity score of the region, is the importance weight of the k0th region, The complexity of building elements in the target area is used to conduct a comprehensive analysis and evaluation of the importance and structural complexity of each area, identify multiple key acceptance areas, and generate key acceptance area data.

[0025] Furthermore, according to the key acceptance area data, considering the spatial positions and environmental obstacles of multiple key areas, the flight path parameters of the UAV are adjusted, the coverage of image acquisition is optimized, and the flight path adjustment parameters are obtained, specifically:

[0026] Analyze and identify the spatial locations and surrounding environmental obstacles of multiple areas, including high-voltage lines and trees, based on the key acceptance area data, and generate environmental obstacle data;

[0027] Based on the environmental obstacle data, calculate the flight path of the UAV and generate a path calculation result;

[0028] According to the path calculation result, the flight parameters of the UAV, including speed, altitude and angle, are adjusted in consideration of coverage requirements and image acquisition quality, and flight path adjustment parameters are generated.

[0029] Furthermore, according to the key acceptance area data, the spatial positions and surrounding environmental obstacles of multiple areas, including high-voltage lines and trees, are analyzed and identified to generate environmental obstacle data, specifically:

[0030] Based on the image data collected by the UAV and the GIS data, the fusion analysis technology is used to identify environmental obstacles; the formula is:

[0031]

[0032] Among them, R represents the environmental obstacle recognition score, w i represents the visibility weight of the i-th obstacle, O i Represents the identification value of the i-th obstacle in the drone image data, T iRepresents the standard position value of the i-th obstacle in the geographic information system, p i Represents the location coefficient of the i-th obstacle and generates environmental obstacle data.

[0033] Furthermore, according to the flight path adjustment parameters, semantic segmentation is performed on the images collected by the drone, and various parts of the distribution network are analyzed and identified to generate distribution network feature recognition data, specifically:

[0034] According to the flight path adjustment parameters, image data of the distribution network area is collected in real time to obtain image collection results;

[0035] Based on the image acquisition results, the ViT algorithm is used to divide the image into multiple fixed-size blocks, and the position encoding is used to convert it into a model input sequence. The multi-head self-attention mechanism and feedforward neural network are combined to analyze the global dependency between image blocks, identify and analyze the geometric and structural features of multiple key distribution network equipment such as transformers, lightning arresters, crossarms and insulators, and generate distribution network image analysis results;

[0036] Based on the distribution network image analysis result, multiple distribution network feature data are analyzed, the geometric shapes of multiple features are calculated and recorded, and distribution network feature recognition data are generated.

[0037] Furthermore, based on the image acquisition results, the ViT algorithm is used to segment the image into multiple fixed-size blocks, and the position encoding is used to convert it into a model input sequence. The global dependency between image blocks is analyzed by combining the multi-head self-attention mechanism and the feedforward neural network, and the geometric and structural features of various key distribution network equipment such as transformers, lightning arresters, crossarms and insulators are identified and analyzed to generate distribution network image analysis results, specifically:

[0038] Based on the image data collected by the drone, the ViT algorithm and spectral analysis technology are used to perform fine-grained identification and analysis of distribution network equipment; the formula is:

[0039]

[0040] Among them, D′ represents the feature difference score, which is used to quantify the recognition accuracy of distribution network equipment such as transformers, lightning arresters, crossarms and insulators in the image, and s i′ represents the spectral sensitivity weight of the i'th device, F i′ represents the feature vector of the i'th device identified by the enhanced ViT algorithm, G i′ represents the predefined device characteristic standard vector, c i′ represents the structural complexity weight of the i'th device, p i′ Represents the pixel ratio of the i'th device in the image, generating the network distribution image analysis result.

[0041] Furthermore, based on the distribution network feature recognition data, by calculating the geometric features of each part of the distribution network, a three-dimensional model of each part of the distribution network is constructed, thereby generating a three-dimensional model of the distribution network, specifically:

[0042] Acquire the distribution network feature recognition data, combine the location information and shooting angle of the drone, perform three-dimensional spatial positioning, and obtain spatial positioning data;

[0043] Based on the spatial positioning data, measure and calculate the three-dimensional geometric parameters of each part of the distribution network, including side length, height and volume, and generate three-dimensional geometric feature data;

[0044] Based on the three-dimensional geometric feature data, three-dimensional structural models of multiple distribution networks are constructed according to geometric features and relative position information to generate a distribution network three-dimensional model.

[0045] Furthermore, based on the spatial positioning data, the three-dimensional geometric parameters of each distribution network part are measured and calculated to generate three-dimensional geometric feature data, specifically:

[0046] Based on the spatial positioning data, stereoscopic vision reconstruction technology is used to perform three-dimensional measurement and calculate the three-dimensional volume of multiple distribution network components; the formula is:

[0047]

[0048] Among them, A p represents the surface area of ​​the pth distribution network part, D p Represents the average depth information related to this distribution network part, They represent the directional coefficients of the x, y, and z axes of the pth distribution network in three-dimensional space, κ represents the normalization coefficient, and V geo Represents three-dimensional geometric volume and generates three-dimensional geometric feature data.

[0049] Furthermore, based on the three-dimensional model of the distribution network, the structural integrity of the distribution network is analyzed, and the impact on the safety of the distribution network is evaluated to generate a structural integrity analysis result, specifically:

[0050] Based on the three-dimensional model of the distribution network, the models of the transformer, the arrester, the cross arm, and the insulator are analyzed and identified, the location and size information of the damage are recorded, and a structural damage identification record is generated;

[0051] Based on the structural damage identification records, a finite element analysis method is used to analyze and calculate the impact of the damaged area on the overall safety of the distribution network, evaluate the structural stability, and generate damage impact analysis data;

[0052] Based on the damage impact analysis data, the overall structural integrity of the distribution network is evaluated, and structural integrity analysis results are generated by considering the identification of defects and safety risks.

[0053] Furthermore, based on the structural integrity analysis results, by comparing with the safety specifications and quality standards of the distribution network acceptance, the matching degree between the distribution network quality and the engineering specifications is evaluated, and the drone engineering acceptance record is generated, specifically:

[0054] Based on the structural integrity analysis results, by comparing with the safety specifications and quality standards for distribution network acceptance, the distribution network structure deviations are identified and recorded, the compliance with multiple standards is evaluated, and safety specification matching data is generated;

[0055] Based on the safety specification matching data, identify and record key structural parameters that do not meet the standards, analyze the target structure and record position information and image data, and generate engineering quality assessment results;

[0056] Based on the project quality assessment results, the matching degree between the distribution network quality and the project specifications is evaluated, the distribution network is accepted, and the drone project acceptance record is generated.

[0057] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a UAV engineering acceptance method based on the ViT detection algorithm, which collects panoramic image data of the distribution network in real time, analyzes the geometric features and structural features of the distribution network, improves the recognition speed and accuracy of key areas in the engineering acceptance process, allows the UAV to adjust the flight path in time during the flight and optimize the coverage of image acquisition, reduces the risk of missing key areas, combines semantic segmentation and three-dimensional modeling, improves the analysis ability of distribution network data, evaluates structural integrity and compares it with the safety specifications and quality standards of the distribution network, generates UAV engineering acceptance records, thereby effectively improving the efficiency, quality and safety of distribution network engineering acceptance. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 is a flow chart of a method implementation of an embodiment of the present invention;

[0059] Figure 2 is a flowchart for implementing step S1 in an embodiment of the present invention;

[0060] Figure 3 is a flowchart for implementing step S2 in an embodiment of the present invention;

[0061] Figure 4 is a flowchart for implementing step S3 in an embodiment of the present invention;

[0062] Figure 5 is a flowchart of the implementation of step S4 in an embodiment of the present invention;

[0063] Figure 6 is a flowchart of the implementation of step S5 in an embodiment of the present invention;

[0064] Figure 7 It is a flow chart for implementing step S6 in an embodiment of the present invention. DETAILED DESCRIPTION

[0065] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0066] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0067] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0068] like Figure 1 As shown, this embodiment provides a method for UAV engineering acceptance based on ViT detection algorithm, including the following steps:

[0069] S1: The image acquisition sensor on the drone collects panoramic image data of the distribution network in real time, analyzes the geometric and structural features of the image, identifies multiple key areas of the distribution network, and obtains key acceptance area data;

[0070] S2: According to the key acceptance area data, considering the spatial positions and environmental obstacles of multiple key areas, adjusting the flight path parameters of the UAV, optimizing the coverage of image acquisition, and obtaining the flight path adjustment parameters;

[0071] S3: adjusting parameters according to the flight path, performing semantic segmentation on the images collected by the drone, analyzing and identifying various parts of the distribution network, and generating distribution network feature recognition data;

[0072] S4: Based on the distribution network feature recognition data, the geometric features of each part of the distribution network are calculated to construct a three-dimensional model of each part of the distribution network, thereby generating a three-dimensional model of the distribution network;

[0073] S5: Based on the three-dimensional model of the distribution network, analyze the structural integrity of the distribution network, evaluate the impact on the safety of the distribution network, and generate a structural integrity analysis result;

[0074] S6: Based on the structural integrity analysis results, by comparing with the safety specifications and quality standards of the distribution network acceptance, the matching degree between the distribution network quality and the engineering specifications is evaluated, and the drone engineering acceptance record is generated.

[0075] Among them, the key acceptance area data include the coordinate positions of multiple areas, the size of the distribution network area, and the distribution network environment characteristics. The flight path adjustment parameters include flight altitude setting parameters, flight speed adjustment records, and flight time prediction data. The distribution network feature identification data includes distribution network size information and structural connection method identification information. The distribution network three-dimensional model includes the three-dimensional geometry of each part of the distribution network, spatial position relationship information, and model data of key load-bearing structures. The structural integrity analysis results include transformer data, lightning arrester data, crossarm data, insulator data, and the type and severity of structural defects. The drone project acceptance record includes distribution network safety rating, specification matching degree, and quality standard achievement percentage information.

[0076] like Figure 2 As shown in the figure, the panoramic image data of the distribution network is collected in real time by the image acquisition sensor on the drone. By analyzing the geometric features and structural features of the image, multiple key areas of the distribution network are identified and the key acceptance area data is obtained, which are as follows:

[0077] S101: Collect image data of the area where the distribution network is located through the image acquisition sensor on the drone, obtain a panoramic image of the area where the distribution network is located, and generate panoramic image data.

[0078] In step S101, the image data of the distribution network area is collected by using an image acquisition sensor and a drone. Multiple flight scans are performed to ensure that high-definition images covering the entire area are obtained from multiple angles. After collecting the image data, the image stitching technology is used to synthesize the single image data into a continuous panoramic image to ensure that there are no gaps in the panoramic image and every detail can be clearly displayed. The image clarity scoring formula is used to evaluate the quality of the synthesized image. The formula is: z =α z ·R z +β z ·G z +γ z ·B z , where α z , β z , γ z are the weighting coefficients of the red, green, and blue channels, respectively, R z , G z , B z The weight coefficient is adjusted to the average brightness value of the corresponding channel to optimize the color balance and contrast of the panoramic image and generate panoramic image data.

[0079] S102: Based on the panoramic image data, analyze the geometric and structural features in the distribution network image, identify the structural boundaries and key construction elements, including transformers, lightning arresters, crossarms, and insulator structures, and generate key construction element data.

[0080] In step S102, based on the panoramic image data, the image is preprocessed, including denoising, contrast enhancement and edge sharpening, to enhance the geometric and structural features in the image. The edge detection algorithm is applied to identify the outline of the structure, and the geometric pattern matching technology is used to identify a variety of key building elements including transformers, lightning arresters, crossarms, and insulator structures. A feature vector is set for each key building element and a database comparison is performed.

[0081] The structure recognition scoring formula is used to evaluate the structure recognition results:

[0082] Among them, S z represents the structure recognition score, is the weight of the i0th key building element, is the matching degree of the key building element in the image, and n0 is the number of types of key building elements, which are used to locate and identify the key building elements in the image and generate key building element data.

[0083] S103: Based on the key construction element data, by evaluating the importance and structural complexity of multiple areas, analyzing and identifying multiple key acceptance areas, and generating key acceptance area data.

[0084] In step S103, the importance of multiple areas is ranked according to the key construction element data, and each area is evaluated using a complexity scoring formula taking into account structural complexity and safety factors;

[0085]

[0086] Among them, C z is the complexity score of the region, is the importance weight of the k0th region, The complexity of building elements in the target area is used to conduct a comprehensive analysis and evaluation of the importance and structural complexity of each area, identify multiple key acceptance areas, and generate key acceptance area data.

[0087] like Figure 3 As shown, according to the key acceptance area data, considering the spatial positions and environmental obstacles of multiple key areas, the flight path parameters of the UAV are adjusted, the coverage of image acquisition is optimized, and the flight path adjustment parameters are obtained, specifically:

[0088] S201: Analyze and identify the spatial locations and surrounding environmental obstacles of multiple areas, including high-voltage lines and trees, based on the key acceptance area data, and generate environmental obstacle data.

[0089] In step S201, the spatial position of each area is mapped according to the key acceptance area data, including the relative position with the surrounding environmental obstacles. The environmental obstacles include high-voltage lines and trees. The positions of the obstacles are fused with the images collected by the drone and the geographic information system data to create an environmental obstacle map. The comprehensive score of environmental obstacle recognition is calculated according to the following formula to generate environmental obstacle data:

[0090]

[0091] Among them, R represents the environmental obstacle recognition score, which represents the quantitative evaluation of the accuracy of each obstacle recognition, and w i Represents the visibility weight of the i-th obstacle, taking into account the recognizability of different types of obstacles in drone images. For example, high-voltage wires and trees have different influences on flight safety. The weights are differentiated to strengthen the recognition priority. i represents the identification value of the i-th obstacle in the UAV image data, providing the actual visual information of the obstacle based on image analysis. i Represents the standard position value of the i-th obstacle in the geographic information system, which is derived from the pre-set GIS data to ensure the accuracy of the geographical location. i Represents the location accuracy coefficient of the i-th type of obstacle, reflects the accuracy of the geographic location data and its weight in the environmental analysis, identifies obstacles around each key acceptance area, and generates environmental obstacle data.

[0092] S202: Calculate the flight path of the UAV based on the environmental obstacle data and generate a path calculation result.

[0093] In step S202, based on the environmental obstacle data, the particle swarm optimization algorithm is used to design the flight path of the UAV. The algorithm first initializes a group of particles, each particle represents a potential flight path solution. In the initial stage, the particles are randomly distributed in the solution space, and each particle has its own position and speed. During the iteration process, the particles update their paths according to the individual best position and the best position found in the group to find the optimal flight route. The fitness evaluation is based on whether the path effectively avoids obstacles such as buildings, high-voltage lines and trees, and the total flight distance of the path. The particles adjust their positions according to the individual best experience and the group best experience. The individual best experience reflects the particle's own historical optimal flight path, and the group best experience reflects the optimal flight path among all particles. The optimal flight path in a complex environment is analyzed, and the optimal solution is gradually approached by continuously adjusting the speed and position of each particle. During the iteration process, the position update of each particle depends on the dynamic relationship between its current position, the individual historical best position and the group historical best position. By iteratively optimizing the flight path, the optimal flight trajectory that minimizes the flight distance and time provides data support for the mission execution of the UAV.

[0094] S203: According to the path calculation result, considering the coverage requirements and image acquisition quality, the flight parameters of the UAV, including speed, altitude and angle, are adjusted to generate flight path adjustment parameters.

[0095] In step S203, according to the path calculation result, considering the coverage requirements and image acquisition quality, the flight parameters of the UAV are adjusted, including adjusting the flight speed, altitude and angle to optimize the image acquisition effect and safety. The flight parameter adjustment formula is:

[0096] F c =λ 1c ·v c +λ 2c ·h c +λ 3c ·θ c

[0097] Among them, F c is the flight parameter adjustment score, λ 1c , 2c , 3c is the adjustment weight of speed, height and angle, v c is the speed, h c is the height, θ c For angles, generate flight path adjustment parameters.

[0098] like Figure 4 As shown, according to the flight path adjustment parameters, the images collected by the drone are semantically segmented, the various parts of the distribution network are analyzed and identified, and the distribution network feature recognition data is generated, specifically:

[0099] S301: According to the flight path adjustment parameters, image data of the area where the distribution network is located is collected in real time to obtain image collection results.

[0100] In step S301, the parameters are adjusted according to the flight path, and the image data of the distribution network area is collected in real time. The UAV adjusts the flight path according to the preset parameters and ensures that the key areas are covered. The quality assessment of the image data adopts the clarity and integrity scoring formula to ensure that the collected image data meets the analysis requirements. The image quality scoring formula is:

[0101] Q b =μ b ·C b +ν b ·D b

[0102] Among them, Q b is the image quality score, μ b and ν bare the weighting factors for clarity and completeness, C b Indicates the clarity score of the image, D b Indicates the data integrity score of the image, ensures that the image acquisition meets the quality standards, and obtains the image acquisition results.

[0103] S302: Based on the image acquisition results, the ViT algorithm is used to divide the image into multiple blocks of fixed size, and the position encoding is used to convert it into a model input sequence. The multi-head self-attention mechanism and feedforward neural network are combined to analyze the global dependencies between image blocks, identify and analyze the geometric and structural features of various key distribution network equipment such as transformers, lightning arresters, crossarms and insulators, and generate distribution network image analysis results.

[0104] In step S302, based on the image acquisition results, the ViT algorithm is applied to analyze the distribution network image data. First, the collected high-resolution images are divided into uniform small blocks, each of which has a fixed size, such as 16x16 pixels. Each small block is converted into a one-dimensional sequence and embedded in a high-dimensional space. The serialized image blocks are processed by the multi-head self-attention mechanism of the ViT model, and the global information is weighted and aggregated in each element of the sequence to strengthen the model's overall understanding of the information in the image. After being processed by a multi-layer Transformer encoder, each layer of the encoder takes the output of the previous layer as input, and further refines the features through the self-attention layer, normalization layer, and feedforward network layer. The model uses the bit The spatial relationship between image blocks is maintained by embedding, ensuring that the converted feature vector retains spatial information. Through multi-level feature extraction and enhancement, the ViT algorithm outputs the feature vector of each image block. The vector set is used to identify and classify key power facilities in the image, such as transformers, poles, and insulators. The output feature vector set is processed by a classification head to map the feature vector to a predefined category label, such as facility type or status. The output result is the classification and location of multiple key components in the distribution network, providing data support for subsequent maintenance work, ensuring the safe operation and efficiency monitoring of the distribution network, and combining spectral analysis technology to perform fine-grained identification and analysis of distribution network equipment. The formula is:

[0105]

[0106] Among them, D′ represents the feature difference score, which is used to quantify the recognition accuracy of distribution network equipment such as transformers, lightning arresters, crossarms and insulators in the image, and s i′ represents the spectral sensitivity weight of the i'th device, F i′ represents the feature vector of the i'th device identified by the enhanced ViT algorithm, G i′ represents the predefined device characteristic standard vector, c i′ represents the structural complexity weight of the i'th device, p i′Represents the pixel ratio of the i'th device in the image.

[0107] S303: Based on the distribution network image analysis result, multiple distribution network feature data are analyzed, geometric shapes of multiple features are calculated and recorded, and distribution network feature recognition data is generated.

[0108] In step S303, based on the distribution network image analysis results, the geometric shape data of each distribution network feature is analyzed. The calculation formula includes the shape, size, aspect ratio and installation position of multiple features in the distribution network. The evaluation formula of the geometric feature is:

[0109]

[0110] Among them, G b is the comprehensive score of geometric features, is the importance weight of the k2th geometric feature, Represents the shape and size scores, It represents the weighted score of the location, p2 is the number of types of geometric features, evaluates and records the geometric characteristics of the distribution network, and generates distribution network feature recognition data.

[0111] like Figure 5 As shown, based on the distribution network feature recognition data, by calculating the geometric features of each part of the distribution network, a three-dimensional model of each part of the distribution network is constructed, thereby generating a three-dimensional model of the distribution network, specifically:

[0112] S401: Acquire the distribution network feature recognition data, combine the position information and shooting angle of the drone, perform three-dimensional spatial positioning, and obtain spatial positioning data.

[0113] In step S401, the distribution network feature recognition data is collected, and the three-dimensional spatial positioning technology is used to locate the drone in combination with the drone's location information and shooting angle. The three-dimensional coordinates of the feature points are calculated using the drone's location and image metadata. The calculation formula for spatial positioning is: o =δ o ·ξ o +η o ·ζ o , where L o is the spatial positioning score for each distribution network feature, δ o and η o are the adjustment coefficients of position and angle, ξ o is the real-time position coordinate of the UAV, ζ o is the angle value of the shooting angle, ensuring the positioning of each distribution network feature in three-dimensional space and obtaining spatial positioning data.

[0114] S402: Based on the spatial positioning data, measure and calculate the three-dimensional geometric parameters of each part of the distribution network, including side length, height and volume, and generate three-dimensional geometric feature data.

[0115] In step S402, based on the spatial positioning data, the three-dimensional geometric parameters of each distribution network part are measured, including the side length, height and volume, and the three-dimensional geometric volumes of multiple components are calculated using three-dimensional modeling technology. The calculation formula of the three-dimensional geometric volume is:

[0116]

[0117] Among them, A p represents the surface area of ​​the pth distribution network part, D p Represents the average depth information related to this distribution network part, They represent the directional coefficients of the x, y, and z axes of the pth distribution network in three-dimensional space, κ represents the normalization coefficient, and V geo Represent the three-dimensional geometric volume, calculate the three-dimensional parameters of multiple distribution network areas, and evaluate the importance of multiple distribution network parts to generate three-dimensional geometric feature data.

[0118] S403: Based on the three-dimensional geometric feature data, according to the geometric features and relative position information, construct three-dimensional structural models of multiple distribution networks to generate a distribution network three-dimensional model.

[0119] In step S403, based on the three-dimensional geometric feature data, a three-dimensional structural model of the distribution network is constructed using modeling technology. The accuracy and practicality of the model are ensured by calculating the geometric features and relative position information. The construction formula of the three-dimensional distribution network model is:

[0120]

[0121] Among them, M o represents the comprehensive evaluation value of the three-dimensional distribution network model, f o It is a comprehensive function based on geometric features and position information. R3 represents the three-dimensional space range. x is the horizontal coordinate in the three-dimensional model, which is used to indicate the horizontal position of the distribution network model. y is the vertical coordinate in the three-dimensional model, which is used to indicate the vertical position of the distribution network model. z is the height coordinate in the three-dimensional model, which is used to indicate the height information of the distribution network model. dx is the differential of x, which is used to calculate the impact of small changes in the horizontal direction on the overall model. dy is the differential of y, which is used to calculate the impact of small changes in the vertical direction on the overall model. dz is the differential of z, which is used to calculate the impact of small changes in the height direction on the overall model. The model is evaluated and constructed to generate a three-dimensional distribution network model.

[0122] like Figure 6As shown, based on the three-dimensional model of the distribution network, the structural integrity of the distribution network is analyzed, and the impact on the safety of the distribution network is evaluated to generate a structural integrity analysis result, which is specifically:

[0123] S501: Based on the three-dimensional model of the distribution network, the models of the transformer, the arrester, the cross arm, and the insulator are analyzed and identified, the location and size information of the damage is recorded, and a structural damage identification record is generated.

[0124] In step S501, according to the three-dimensional distribution network model, image analysis is used to detect and identify the damaged areas of transformers, lightning arresters, crossarms, and insulators in the distribution network. The visualization effect of the distribution network is optimized by automatic image enhancement technology. Then, a deep learning model is used to identify and classify differentiated types of structural damage. To quantify the damage, the following formula is used:

[0125]

[0126] Among them, H p represents the total damage quantification value, is the severity factor assigned according to the injury type, Represents the maximum linear size of the damage, n4 is the number of damage types, which improves the accuracy of damage detection. By quantifying the size, the location and size of each damage are recorded to generate a structural damage identification record.

[0127] S502: Based on the structural damage identification record, a finite element analysis method is used to analyze and calculate the impact of the damaged area on the overall safety of the distribution network, evaluate the structural stability, and generate damage impact analysis data.

[0128] In step S502, based on the structural damage identification record, the finite element analysis method is used to evaluate the impact of each damaged area on the overall safety of the distribution network, considering the impact of crack width, corrosion depth, and structural defects on the load bearing capacity and durability of the distribution network. By integrating the impact of multiple damaged areas, the overall stability of the structure is evaluated and damage impact analysis data is generated.

[0129] Finite element analysis method, according to the formula:

[0130]

[0131] Calculate the stress distribution of the distribution network structure, where σ is stress, which is used to indicate the magnitude of the force generated by external forces inside the material, E is the elastic modulus of the distribution network material, which is used to describe the material's ability to resist deformation, ∈ is strain, which is used to describe the degree of deformation per unit length of the material, and S f is the crack shape factor, which is used to adjust the effect of crack shape on stress concentration. eis the corrosion environment factor, which is used to adjust the impact of differentiated environment on material properties. ν is the Poisson's ratio, which is used to describe the proportional relationship between the lateral and longitudinal deformations of the material after being subjected to force. k is the correction coefficient, which is used to comprehensively consider the influence of material properties and structural characteristics to ensure the accuracy of the calculation results.

[0132] The specific execution process of the formula is as follows:

[0133] Through detailed inspection of the damaged area, the crack width and corrosion depth data are obtained, and the crack shape factor S is calculated based on the environment and crack morphology analysis. f and corrosion environment factor C e , use material test data to determine the elastic modulus E and Poisson's ratio ν, adjust the value of the correction factor k based on historical data, substitute the parameters into the finite element analysis formula, and calculate the stress σ, which is used to evaluate the overall stability of the structure and guide the implementation of maintenance or reinforcement measures.

[0134] S503: Based on the damage impact analysis data, the overall structural integrity of the distribution network is evaluated, and identification defects and safety risks are considered to generate a structural integrity analysis result.

[0135] In step S503, the structural integrity assessment is performed based on the damage impact analysis data. By integrating the damage data from micro to macro, the impact on the distribution network function and safety is evaluated, and the structural integrity is quantified using a composite risk scoring method. The formula is:

[0136]

[0137] Among them, J p represents the structural integrity risk index, is the risk factor of each injury, is the quantified value of damage severity, q4 is the total number of damages, and e is the base of the natural logarithm. By considering the cumulative effect of the identified damages, the integrity of the target distribution network structure is evaluated and the structural integrity analysis results are generated.

[0138] like Figure 7 As shown, according to the structural integrity analysis results, by comparing with the safety specifications and quality standards of distribution network acceptance, the matching degree between distribution network quality and engineering specifications is evaluated, and the drone engineering acceptance record is generated, specifically:

[0139] S601: Based on the structural integrity analysis results, by comparing with the safety specifications and quality standards of the distribution network acceptance, the distribution network structure deviations are identified and recorded, the compliance with multiple standards is evaluated, and safety specification matching data is generated.

[0140] In step S601, based on the structural integrity analysis results, the safety specifications and quality standards of the distribution network are compared to verify whether the distribution network structure meets the relevant regulations. The safety specification matching analysis is performed using the formula:

[0141]

[0142] Among them, C g represents the safety specification matching score, is the importance weight of the i5th criterion, is the actual measured value, is the standard value, n5 is the number of standard items checked, and by comparing the deviation between the actual value and the standard value, the compliance degree of the distribution network structure is evaluated and recorded to generate safety specification matching data.

[0143] S602: Based on the safety specification matching data, identify and record key structural parameters that do not meet the standards, analyze the target structure and record position information and image data, and generate engineering quality assessment results.

[0144] In step S602, based on the safety specification matching data, the key structural parameters that do not meet the standards are identified and recorded, including analyzing the target key parameters, determining the location and coordinating with the image data collected by the drone. The analysis and recording process uses the following calculation model:

[0145]

[0146] Among them, P g The score for the distribution network quality assessment is is the risk factor of the unqualified parameter, is the measured value, is the standard requirement value, m5 is the number of parameters that do not meet the requirements, and the project quality assessment results are generated by recording the deviations of key structural parameters and evaluating the impact on the overall project quality.

[0147] S603: Based on the project quality assessment result, the matching degree between the distribution network quality and the project specification is assessed, the distribution network is accepted, and a drone project acceptance record is generated.

[0148] In step S603, based on the engineering quality assessment results, the distribution network quality and specification matching degree are evaluated, the distribution network acceptance is performed, and the overall matching degree is calculated. The formula is:

[0149]

[0150] Among them, V g For project acceptance matching, is the adjustment coefficient, reflecting the impact of multiple unqualified parameters on project acceptance. is the degree of deviation of multiple parameters, q5 is the total number of parameters evaluated, and e is the base of the natural logarithm. The impact of each unqualified parameter is quantified through the exponential decay model, which provides a data basis for the acceptance of the entire distribution network and generates the drone project acceptance record.

[0151] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0152] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0153] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0154] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0155] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

Claims

1. A UAV project acceptance method based on ViT detection algorithm, characterized in that: include: The image acquisition sensor on the drone collects panoramic image data of the distribution network in real time, and analyzes the geometric and structural features of the image to identify multiple key areas of the distribution network and obtain key acceptance area data; According to the key acceptance area data, considering the spatial positions and environmental obstacles of multiple key areas, adjusting the flight path parameters of the UAV, optimizing the coverage of image acquisition, and obtaining the flight path adjustment parameters; Adjusting parameters according to the flight path, performing semantic segmentation on the images collected by the drone, analyzing and identifying various parts of the distribution network, and generating distribution network feature recognition data; Based on the distribution network feature recognition data, by calculating the geometric features of each part of the distribution network, a three-dimensional model of each part of the distribution network is constructed, thereby generating a three-dimensional model of the distribution network; Based on the three-dimensional model of the distribution network, the structural integrity of the distribution network is analyzed, and the impact on the safety of the distribution network is evaluated to generate a structural integrity analysis result; According to the structural integrity analysis results, by comparing with the safety specifications and quality standards of distribution network acceptance, the matching degree between distribution network quality and engineering specifications is evaluated, and the UAV engineering acceptance record is generated.

2. The method for acceptance of UAV engineering based on ViT detection algorithm according to claim 1 is characterized in that: The image acquisition sensor on the drone collects the panoramic image data of the distribution network in real time. By analyzing the geometric and structural features of the image, multiple key areas of the distribution network are identified and the key acceptance area data is obtained, specifically: The image data of the distribution network area is collected by the image acquisition sensor on the drone, a panoramic image of the distribution network area is obtained, and panoramic image data is generated; Based on the panoramic image data, analyzing geometric and structural features in the distribution network image, identifying structural boundaries and key construction elements, and generating key construction element data; Based on the key construction element data, by evaluating the importance and structural complexity of multiple areas, multiple key acceptance areas are analyzed and identified to generate key acceptance area data.

3. The method for acceptance of UAV engineering based on ViT detection algorithm according to claim 2 is characterized in that: Based on the panoramic image data, the geometric and structural features in the distribution network image are analyzed to identify the structural boundaries and key building elements, including beams, columns and supporting structures, and generate key building element data, specifically: Based on the panoramic image data, the image is preprocessed to enhance the geometric and structural features in the image; Apply edge detection algorithms to identify the outline of the structure, use geometric pattern matching technology to identify multiple key building elements, set feature vectors for each key building element and perform database comparison; The structure recognition scoring formula is used to evaluate the structure recognition results: Among them, S z represents the structure recognition score, is the weight of the i0th key building element, is the matching degree of the key building element in the image, and n0 is the number of types of key building elements, which are used to locate and identify the key building elements in the image and generate key building element data; According to the key construction element data, by evaluating the importance and structural complexity of multiple areas, multiple key acceptance areas are analyzed and identified, and key acceptance area data is generated, specifically: According to the key construction element data, the importance of multiple areas is ranked, and each area is evaluated using a complexity scoring formula considering structural complexity and safety factors; Among them, C z is the complexity score of the region, is the importance weight of the k0th region, The complexity of building elements in the target area is used to conduct a comprehensive analysis and evaluation of the importance and structural complexity of each area, identify multiple key acceptance areas, and generate key acceptance area data.

4. The method for acceptance of UAV engineering based on ViT detection algorithm according to claim 1 is characterized in that: According to the key acceptance area data, considering the spatial positions and environmental obstacles of multiple key areas, the flight path parameters of the UAV are adjusted, the coverage of image acquisition is optimized, and the flight path adjustment parameters are obtained, specifically: Analyze and identify the spatial locations and surrounding environmental obstacles of multiple areas, including high-voltage lines and trees, based on the key acceptance area data, and generate environmental obstacle data; Based on the environmental obstacle data, calculate the flight path of the UAV and generate a path calculation result; According to the path calculation result, the flight parameters of the UAV, including speed, altitude and angle, are adjusted in consideration of coverage requirements and image acquisition quality, and flight path adjustment parameters are generated.

5. The method for acceptance of UAV engineering based on ViT detection algorithm according to claim 4 is characterized in that: According to the key acceptance area data, the spatial locations and surrounding environmental obstacles of multiple areas, including high-voltage lines and trees, are analyzed and identified, and environmental obstacle data is generated, specifically: Based on the image data collected by the UAV and the GIS data, the fusion analysis technology is used to identify environmental obstacles; the formula is: Among them, R represents the environmental obstacle recognition score, w i represents the visibility weight of the i-th obstacle, O i Represents the identification value of the i-th obstacle in the drone image data, T i Represents the standard position value of the i-th obstacle in the geographic information system, p i Represents the location coefficient of the i-th obstacle and generates environmental obstacle data.

6. The method for acceptance of UAV engineering based on ViT detection algorithm according to claim 1, characterized in that: According to the flight path adjustment parameters, the images collected by the drone are semantically segmented, the various parts of the distribution network are analyzed and identified, and the distribution network feature recognition data is generated, specifically: According to the flight path adjustment parameters, image data of the distribution network area is collected in real time to obtain image collection results; Based on the image acquisition results, the ViT algorithm is used to divide the image into multiple fixed-size blocks, and the position encoding is used to convert it into a model input sequence. The multi-head self-attention mechanism and feedforward neural network are combined to analyze the global dependency between image blocks, identify and analyze the geometric and structural features of multiple key distribution network equipment such as transformers, lightning arresters, crossarms and insulators, and generate distribution network image analysis results; Based on the distribution network image analysis result, multiple distribution network feature data are analyzed, the geometric shapes of multiple features are calculated and recorded, and distribution network feature recognition data are generated.

7. The method for acceptance of UAV engineering based on ViT detection algorithm according to claim 6 is characterized in that: Based on the image acquisition results, the ViT algorithm is used to divide the image into multiple fixed-size blocks, and the position encoding is used to convert it into a model input sequence. The multi-head self-attention mechanism and feedforward neural network are combined to analyze the global dependencies between image blocks, identify and analyze the geometric and structural features of various key distribution network equipment such as transformers, lightning arresters, crossarms and insulators, and generate distribution network image analysis results, specifically: Based on the image data collected by the drone, the ViT algorithm and spectral analysis technology are used to perform fine-grained identification and analysis of distribution network equipment; the formula is: Among them, D′ represents the feature difference score, which is used to quantify the recognition accuracy of distribution network equipment such as transformers, lightning arresters, crossarms and insulators in the image, and s i′ represents the spectral sensitivity weight of the i'th device, F i′ represents the feature vector of the i'th device identified by the enhanced ViT algorithm, G i′ represents the predefined device characteristic standard vector, c i′ represents the structural complexity weight of the i'th device, p i′ Represents the pixel ratio of the i'th device in the image, generating the network distribution image analysis result.

8. The method for acceptance of UAV engineering based on ViT detection algorithm according to claim 1, characterized in that: Based on the distribution network feature recognition data, the geometric features of each part of the distribution network are calculated to construct a three-dimensional model of each part of the distribution network, thereby generating a three-dimensional model of the distribution network, specifically: Acquire the distribution network feature recognition data, combine the location information and shooting angle of the drone, perform three-dimensional spatial positioning, and obtain spatial positioning data; Based on the spatial positioning data, measure and calculate the three-dimensional geometric parameters of each part of the distribution network, including side length, height and volume, and generate three-dimensional geometric feature data; Based on the three-dimensional geometric feature data, three-dimensional structural models of multiple distribution networks are constructed according to geometric features and relative position information to generate a distribution network three-dimensional model.

9. The method for acceptance of UAV engineering based on ViT detection algorithm according to claim 8, characterized in that: Based on the spatial positioning data, the three-dimensional geometric parameters of each distribution network part are measured and calculated to generate three-dimensional geometric feature data, specifically: Based on the spatial positioning data, stereoscopic vision reconstruction technology is used to perform three-dimensional measurement and calculate the three-dimensional volume of multiple distribution network components; the formula is: Among them, A p represents the surface area of ​​the pth distribution network part, D p Represents the average depth information related to this distribution network part, They represent the directional coefficients of the x, y, and z axes of the pth distribution network in three-dimensional space, κ represents the normalization coefficient, and V geo Represents three-dimensional geometric volume and generates three-dimensional geometric feature data.

10. The method for acceptance of UAV engineering based on ViT detection algorithm according to claim 1, characterized in that: Based on the three-dimensional model of the distribution network, the structural integrity of the distribution network is analyzed, and the impact on the safety of the distribution network is evaluated to generate structural integrity analysis results, specifically: Based on the three-dimensional model of the distribution network, the models of the transformer, the arrester, the cross arm, and the insulator are analyzed and identified, the location and size information of the damage are recorded, and a structural damage identification record is generated; Based on the structural damage identification records, a finite element analysis method is used to analyze and calculate the impact of the damaged area on the overall safety of the distribution network, evaluate the structural stability, and generate damage impact analysis data; Based on the damage impact analysis data, the overall structural integrity of the distribution network is evaluated, and structural integrity analysis results are generated by considering the identification of defects and safety risks.

11. The method for acceptance of UAV engineering based on ViT detection algorithm according to claim 1, characterized in that: According to the structural integrity analysis results, by comparing with the safety specifications and quality standards of distribution network acceptance, the matching degree between distribution network quality and engineering specifications is evaluated, and the drone engineering acceptance record is generated, specifically: Based on the structural integrity analysis results, by comparing with the safety specifications and quality standards for distribution network acceptance, the distribution network structure deviations are identified and recorded, the compliance with multiple standards is evaluated, and safety specification matching data is generated; Based on the safety specification matching data, identify and record key structural parameters that do not meet the standards, analyze the target structure and record position information and image data, and generate engineering quality assessment results; Based on the project quality assessment results, the matching degree between the distribution network quality and the project specifications is evaluated, the distribution network is accepted, and the drone project acceptance record is generated.

Images