Distribution network unmanned aerial vehicle acceptance method based on visual fusion tracking and image recognition

Abstract

The invention relates to a distribution network unmanned aerial vehicle acceptance method based on visual fusion tracking and image recognition, and relates to the technical field of distribution network project acceptance, and the method comprises the steps: employing a visual fusion tracking technology, fusing multi-dimensional real-time visual information, and combining a target centering algorithm based on an initial unmanned aerial vehicle inspection route; real-time identification and continuous tracking of the power distribution equipment are realized; in the tracking process, the flight height and course of the unmanned aerial vehicle are dynamically adjusted through an AI autonomous obstacle avoidance model, and a distribution network facility image of the power distribution equipment is collected; extracting construction process defect features and equipment nameplate identification features from the processed distribution network facility image; and generating and outputting an acceptance report according to the grading result of the construction process defects and the verification result of the equipment model, by adopting the technical scheme, manual intervention is not needed, the acceptance period is shortened, the efficiency is improved, and misjudgment and leak detection caused by poor image quality are reduced, so that the requirements of rapid and accurate acceptance of large-scale distribution network engineering are met.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of distribution network engineering acceptance, in particular to a distribution network unmanned aerial vehicle acceptance method based on visual fusion tracking and image recognition. BACKGROUND

[0002] With the rapid development of China's electric power industry, as the key link connecting the main network and users, the construction scale of distribution network is increasing, and the requirements for construction quality and operation reliability are also increasing. The acceptance work of 10kV and below distribution network overhead line is the first line of defense to ensure the safe and stable operation of the power grid. The traditional acceptance method mainly relies on manual pole inspection or ground telescope observation, which has many disadvantages such as high labor intensity, high operation risk, low efficiency, and blind area in inspection.

[0003] In order to overcome the defects of traditional manual acceptance, the unmanned aerial vehicle technology can be used for distribution network acceptance. Specifically, the unmanned aerial vehicle equipped with high-definition cameras, visual sensors and flight control systems can be used to conduct aerial inspection and acceptance of the construction quality and equipment status of overhead lines in combination with geographic information (GIS) and engineering account data.

[0004] However, most of the existing unmanned aerial vehicles for acceptance are only used as basic aerial image acquisition platforms, and their onboard systems lack intelligent real-time processing and analysis capabilities. A large amount of high-definition image and video data collected by the unmanned aerial vehicle during the inspection process needs to be transmitted to the ground workstation in full. The subsequent defect identification work completely relies on manual judgment and screening by acceptance personnel. This "machine collection and human judgment" mode not only consumes a lot of manpower and time, resulting in long acceptance period and low efficiency, but also easily causes visual fatigue due to human factors, leading to missed detection or misjudgment, which is difficult to meet the demand for rapid and accurate acceptance of large-scale distribution network engineering. SUMMARY

[0005] Therefore, the present application provides a distribution network unmanned aerial vehicle acceptance method based on visual fusion tracking and image recognition, which mainly aims to solve the technical problem that the existing unmanned aerial vehicles for acceptance mostly adopt the "machine collection and human judgment" mode, which not only consumes a lot of manpower and time, resulting in long acceptance period and low efficiency, but also easily causes visual fatigue due to human factors, leading to missed detection or misjudgment, which is difficult to meet the demand for rapid and accurate acceptance of large-scale distribution network engineering.

[0006] In a first aspect, the present application provides a distribution network unmanned aerial vehicle acceptance method, comprising: generating an initial unmanned aerial vehicle inspection route based on the engineering account and geographic information data of the distribution network overhead line; based on the initial unmanned aerial vehicle inspection route, using visual fusion tracking technology to fuse multi-dimensional real-time visual information, and combining with a target centering algorithm to realize real-time identification and continuous tracking of distribution equipment; In the tracking process, the multi-dimensional real-time visual information is combined, and an AI autonomous obstacle avoidance model is used to dynamically adjust the flight height and direction of the unmanned aerial vehicle to collect power distribution facility images of the power distribution equipment; The power distribution facility images are processed by an on-board edge computing intelligent terminal, and image recognition technology is applied to extract construction process defect features and equipment nameplate identification features from the processed power distribution facility images; The construction process defect features are compared with a preset standard library to identify and grade the construction process defects of the power distribution equipment, and the equipment model of the power distribution equipment is identified and obtained according to the extracted equipment nameplate identification features, and the equipment model is checked with the equipment model in the engineering account; According to the grading result of the construction process defects and the checking result of the equipment model, an acceptance report is generated and output.

[0007] In a second aspect, the present application provides a power distribution unmanned aerial vehicle acceptance device, comprising: A generation module is configured to generate an initial unmanned aerial vehicle inspection route based on an engineering account and geographic information data of a power distribution overhead line; A tracking module is configured to use a visual fusion tracking technology to fuse multi-dimensional real-time visual information, and combine a target centering algorithm to realize real-time identification and continuous tracking of power distribution equipment based on the initial unmanned aerial vehicle inspection route; A collection module is configured to combine the multi-dimensional real-time visual information in the tracking process, and use an AI autonomous obstacle avoidance model to dynamically adjust the flight height and direction of the unmanned aerial vehicle to collect power distribution facility images of the power distribution equipment; An extraction module is configured to process the power distribution facility images by an on-board edge computing intelligent terminal, and apply image recognition technology to extract construction process defect features and equipment nameplate identification features from the processed power distribution facility images; An identification module is configured to compare the construction process defect features with a preset standard library to identify and grade the construction process defects of the power distribution equipment, and identify and obtain the equipment model of the power distribution equipment according to the extracted equipment nameplate identification features, and check the equipment model with the equipment model in the engineering account; An output module is configured to generate and output an acceptance report according to the grading result of the construction process defects and the checking result of the equipment model.

[0008] In a third aspect, the present application provides an electronic device, comprising a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, wherein the processor executes the computer program to realize the power distribution unmanned aerial vehicle acceptance method of the first aspect.

[0009] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the power distribution network drone acceptance method described in the first aspect. By employing the above technical solution, this application provides a method for acceptance testing of power distribution network drones based on visual fusion tracking and image recognition. Compared with existing technologies, this application can generate an initial drone inspection route based on engineering ledgers and geographic information data of overhead power distribution lines. Based on the initial drone inspection route, visual fusion tracking technology is used to integrate multi-dimensional real-time visual information and combine it with a target centering algorithm to achieve real-time identification and continuous tracking of power distribution equipment. During the tracking process, the drone's flight altitude and heading are dynamically adjusted through an AI autonomous obstacle avoidance model, combined with multi-dimensional real-time visual information, to collect images of power distribution facilities. The images of power distribution facilities are processed by an airborne edge computing intelligent terminal, and image recognition technology is applied to extract construction process defect features and equipment nameplate identification features from the processed images. The construction process defect features are compared with a preset standard library to identify and classify the construction process defects of the power distribution equipment. At the same time, based on the extracted equipment nameplate identification features, the equipment model of the power distribution equipment is identified and verified against the equipment model in the engineering ledger. Based on the classification results of the construction process defects and the verification results of the equipment model, an acceptance report is generated and output. By adopting the above technical solution, this application automatically generates initial inspection routes covering all power distribution equipment based on engineering ledgers and geographic information data, eliminating the need for manual on-site surveys and reducing fieldwork workload by more than 50%. From route planning and construction inspection to acceptance report generation, no manual intervention is required throughout the entire process. The drone automatically performs data collection, analysis, and decision-making tasks, with data transmitted back to the cloud in real time, forming a closed loop of "collection-processing-decision-feedback," reducing the acceptance cycle and improving efficiency.

[0010] This application utilizes visual fusion tracking technology, combined with a target centering algorithm, to ensure that the drone continuously locks onto power distribution equipment in complex environments (such as strong winds or obstructions), avoiding repeated inspections due to target loss. Combined with an AI-powered autonomous obstacle avoidance model, the drone can safely and flexibly approach the equipment, avoiding image blurring caused by excessive distance or angular deviations. This ensures data quality from the source, reducing misjudgments and missed detections due to poor image quality, thus meeting the needs of rapid and accurate acceptance testing for large-scale power distribution network projects.

[0011] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below. Attached Figure Description

[0012] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0013] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 A flowchart illustrating a method for accepting power distribution network drones, provided as an embodiment of this application; Figure 2 This is a schematic diagram of the structure of a power distribution network drone acceptance device provided in an embodiment of this application. Detailed Implementation

[0015] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of this application, including various details to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0016] The following describes, with reference to the accompanying drawings, an embodiment of the present application of a method for accepting power distribution network drones based on visual fusion tracking and image recognition.

[0017] This application provides a method for acceptance testing of power distribution network drones based on visual fusion tracking and image recognition. The main purpose is to solve the technical problems that existing drones used for acceptance testing mostly adopt the "machine-based inspection and human judgment" mode, which not only consumes a lot of manpower and time, resulting in long acceptance cycles and low efficiency, but is also prone to visual fatigue caused by human factors, leading to missed inspections or misjudgments, making it difficult to meet the technical needs of rapid and accurate acceptance testing for large-scale power distribution network projects.

[0018] like Figure 1 As shown, an embodiment of this application provides a method for accepting power distribution network drones, including: Step 101: Generate the initial UAV inspection route based on the engineering ledger and geographic information data of the overhead power distribution line.

[0019] For the embodiments of this application, the engineering ledger and geographic information data (GIS data) of the overhead power distribution lines to be inspected can be obtained first. The engineering ledger may contain structured data such as tower coordinates, equipment models, and design parameters, while the geographic information data provides high-precision geospatial information of the line corridor, such as terrain elevation and obstacle distribution. By fusing these two types of data, an initial UAV inspection route covering all equipment to be inspected, conforming to flight safety regulations, and with optimal path can be automatically planned. This initial UAV inspection route may include latitude, longitude, altitude, and speed parameters. The initial UAV inspection route can be loaded into the UAV's flight control system as a reference path for autonomous flight.

[0020] Step 102: Based on the initial UAV inspection route, visual fusion tracking technology is adopted to integrate multi-dimensional real-time visual information and combine it with target centering algorithm to achieve real-time identification and continuous tracking of power distribution equipment.

[0021] To ensure the drone can accurately aim at and photograph the target device, this embodiment employs advanced visual fusion tracking technology. This step specifically includes the following sub-steps: Based on the initial drone inspection route, images of the power distribution area of ​​the power distribution equipment are collected using the drone's visible light camera; The relative distance and angle data between the UAV and the current visual target are collected synchronously by the UAV's visual ranging module, and the power distribution area image is fused with the relative distance and angle data to generate fused multi-dimensional real-time visual information. Based on the fused multi-dimensional real-time visual information, the pre-trained target detection model is invoked to identify the pixel regions of power distribution equipment in the power distribution area image. The pixel coordinates of the power distribution equipment are locked and recorded by the target box, so as to determine the physical center coordinates of the power distribution area image based on the pixel coordinates. The target centering algorithm calculates the coordinate offset between the center coordinates of the target box and the physical center coordinates, and dynamically adjusts the flight attitude of the UAV based on the coordinate offset to keep the target box centered in the image, thereby enabling real-time identification and continuous tracking of power distribution equipment.

[0022] In this embodiment, the drone can fly along the initial drone inspection route and use its onboard visible light camera (with a resolution of up to 4K) to collect high-definition images of the power distribution equipment area in front in real time.

[0023] Simultaneously, the relative distance (accuracy ±0.1m) and angle data (pitch angle ±15°, yaw angle ±10°) between the UAV and the current visual target (such as insulators, clamps, etc.) can be simultaneously collected via the visual ranging module on the UAV (such as binocular vision or structured light cameras). The relative distance and angle data can be sent to the onboard edge computing terminal along with the image of the power distribution equipment area for spatiotemporal alignment and data fusion processing, generating multi-dimensional real-time visual information containing two-dimensional image information and three-dimensional spatial information.

[0024] Based on the fused multi-dimensional real-time visual information, a pre-trained target detection model (such as the YOLO series model) can be invoked to perform target detection on each frame of the image, thereby identifying the pixel region of the power distribution equipment. At the same time, the identified power distribution equipment can be located in the image by the target box (such as the insulator target box size ≥ 50×50 pixels) and the pixel coordinates of the target box can be recorded, such as the upper left corner coordinates (x1, y1) and the lower right corner coordinates (x2, y2).

[0025] To achieve continuous tracking, the drone's flight attitude can be dynamically adjusted using a target centering algorithm. The specific calculation process is as follows: First, the center coordinates of the target box in the current frame ((x1+x2) / 2, (y1+y2) / 2) can be calculated based on the pixel coordinates.

[0026] The coordinate offset is obtained by comparing the center coordinates of the target bounding box with the physical center coordinates of the image (e.g., for an image with a resolution of 1920×1080, its physical center is (960, 540)). The coordinate offset may include a lateral offset. and vertical offset .

[0027] Coordinate offsets can be converted into UAV flight control commands: for example, if If the value is >0 (target is slightly to the right in the image), then output the command "Slightly adjust course to the left"; if... If the value is >0 (the target is lower in the image), then the command "adjust height upwards" will be output.

[0028] At the same time, visual ranging data can be combined to output a "fly backward" command if the drone is too close to the target (for example, the drone is less than 5 meters away from the target) to ensure that the equipment image is clear and the size is appropriate.

[0029] Through the aforementioned closed-loop control, the drone can always lock onto the target device, laying the foundation for subsequent high-quality image acquisition.

[0030] Step 103: During the tracking process, the drone's flight altitude and heading are dynamically adjusted using an AI autonomous obstacle avoidance model, in conjunction with multi-dimensional real-time visual information, in order to collect images of the power distribution equipment's network facilities.

[0031] In complex power distribution network environments, drones need to possess autonomous obstacle avoidance capabilities. This embodiment can achieve this function through an AI-powered autonomous obstacle avoidance model, specifically including: The obstacle recognition model is invoked to perform data analysis on multi-dimensional real-time visual information and identify obstacles in the environment, including obstacle type and obstacle size; By using visual ranging data and lidar data, the actual distance between the obstacle and the drone is calculated, as well as the vertical height difference between the top of the obstacle and the power line. Based on the actual distance value and / or vertical height difference, the risk level of the obstacle is determined, and obstacle avoidance instructions are output according to the risk level. The drone's flight altitude and heading are dynamically adjusted according to the obstacle avoidance instructions through the AI ​​autonomous obstacle avoidance model, and images of power distribution equipment and power distribution facilities are collected.

[0032] Determining the risk level of an obstacle based on the actual distance value and / or the difference in vertical height may include: If the actual distance value and / or vertical height difference is less than the minimum preset threshold, or if the obstacle intrudes into the inspection route range, the risk level of the obstacle is determined to be high risk level. If the actual distance value and / or vertical height difference is greater than or equal to the minimum preset threshold, and the actual distance value and / or vertical height difference is less than the maximum preset threshold, and the obstacle does not intrude into the inspection route range, the risk level of the obstacle is determined to be medium risk level. If the actual distance value and / or vertical height difference is greater than or equal to the maximum preset threshold, the obstacle's risk level is determined to be low risk.

[0033] In this embodiment, a pre-trained obstacle recognition model can be invoked to continuously analyze the multi-dimensional real-time visual information fused in step 102, so as to perform target detection on the fused environmental image, identify the types of obstacles in the environment (such as trees, buildings, other lines, etc.) and determine the size of the obstacles (such as length, width and height).

[0034] The actual distance L between the obstacle and the drone, as well as the vertical height difference H between the top of the obstacle and the power line, can be accurately calculated using visual ranging data and LiDAR data. Risk levels can then be categorized based on the actual distance and vertical height difference. For example, if L < 3m or H < 1.5m (safety threshold), or if an obstacle intrudes into the preset inspection route range, the obstacle's risk level is determined to be high risk. If 3m≤L or H<5m (safety threshold), and the obstacle does not intrude into the preset inspection route range but is relatively close, then the risk level of the obstacle is determined to be medium risk level; If L or H ≥ 5m (safety threshold), it indicates no interference with the flight path, and the obstacle risk level is determined to be low risk level.

[0035] Based on the assessed risk level, the AI-powered obstacle avoidance model outputs corresponding obstacle avoidance commands (such as "increase altitude" and "yaw left / right"), dynamically adjusting the drone's flight path (such as flight altitude and heading). After completing obstacle avoidance, the drone can automatically return to its original inspection route and continue performing the tracking and image acquisition tasks in step 102, collecting images of the power distribution equipment's network facilities.

[0036] Step 104: Process the images of the power distribution network facilities through the airborne edge computing intelligent terminal, and apply image recognition technology to extract construction process defect features and equipment nameplate identification features from the processed images of the power distribution network facilities.

[0037] One of the core innovations of this embodiment is that it places complex image analysis tasks on the airborne end for real-time processing, which greatly improves efficiency.

[0038] For embodiments of this application, extracting construction process defect features from the processed power distribution network facility image may include: The images of power distribution equipment in the images of power distribution network facilities are segmented using a lightweight semantic segmentation model, and defective regions in the images of power distribution equipment are located by threshold filtering. Extract the geometric and surface texture features of the defect area to obtain the construction process defect features; The geometric features include the area of ​​the defect region calculated based on the number of pixels within the defect region and the image pixel resolution; and / or, the roundness calculated based on the area of ​​the defect region; the surface texture features include contrast and / or correlation calculated based on the gray-level co-occurrence matrix.

[0039] In this embodiment, in order to extract the features of construction process defects, the collected images (i.e., power distribution facility images) are first segmented using a lightweight semantic segmentation model (such as MobileNetV3-Seg) to segment the power distribution equipment images. The power distribution equipment images can accurately outline the contour images of power distribution equipment such as insulators and conductors.

[0040] Secondly, suspected defect areas can be located within the segmented equipment area using methods such as threshold screening. Specifically, taking insulator damage as an example, the grayscale variance of the segmented insulator area can be used... The gray-scale variance of the normal region of the insulator pre-stored in the model If a comparison is made, >1.5 × If so, it is determined to be a suspected damaged area.

[0041] Finally, construction process defect features of suspected damaged areas can be extracted. These features may include geometric features and surface texture features.

[0042] Geometric features can be defined as: counting the number of pixels within the suspected damaged area. Combined with image pixel resolution Calculate the area of ​​the defect region The formula is as follows:

[0043] In the formula, Area of ​​the defect region, unit: , This represents the number of pixels within the damaged area. Pixel resolution.

[0044] Calculate the circularity of the defect area based on its area. The formula is as follows:

[0045] In the formula, The roundness of the defect area, This represents the area of ​​the defective region. Let be the perimeter of the defective region.

[0046] Among them, normal insulators have no holes. ≈0, while the C value of damaged holes is usually in the range of 0. Within the range.

[0047] Surface texture features can be calculated by the gray-level co-occurrence matrix. (i,j are gray levels, step size=1, angle=0°), extract texture features such as contrast and correlation of defect areas; The contrast ratio formula is as follows:

[0048] In the formula, For the contrast of the defective area, This is the gray-level co-occurrence matrix.

[0049] The correlation formula is as follows:

[0050] In the formula, For the correlation of defective areas, The gray-level co-occurrence matrix, The standard deviation related to gray level i, is the standard deviation related to gray level j.

[0051] Typically, defective areas have higher contrast and lower correlation.

[0052] In addition, for defects such as loose conductors, the contour point set after conductor segmentation can be calculated. Distance between adjacent points If it exists ( If the distance is the average distance, then it is determined to be a suspected loose area.

[0053] For embodiments of this application, extracting equipment nameplate identification features from the processed distribution network facility image may include: Detect the nameplate area in images of power distribution network facilities; After filtering out false targets in the nameplate area by contour rectangle, the geometric and surface texture features of the characters in the nameplate area are extracted. Then, optical character recognition technology is used to recognize the characters based on the geometric and surface texture features, and the recognition results are used as the identification features of the equipment nameplate.

[0054] In this embodiment, in order to extract the identification features of the equipment nameplate, the YOLOv8-nano model can be used to quickly detect the nameplate area in the image (i.e., the image of the distribution network facility).

[0055] Then, irregularly shaped false targets (such as stains, scratches, etc.) can be filtered out by contour rectangle.

[0056] Finally, the geometric and surface texture features of the characters within the nameplate area can be extracted, and optical character recognition (OCR) technology can be used to recognize the characters. The recognized text information (such as equipment model and serial number) is used as the equipment nameplate identification feature.

[0057] Step 105: Compare the construction process defect characteristics with the preset standard library to identify and classify the construction process defects of the power distribution equipment. At the same time, based on the extracted equipment nameplate identification characteristics, identify and obtain the equipment model of the power distribution equipment, and verify the model with the equipment model in the project ledger.

[0058] In this embodiment, the airborne terminal can match the extracted construction process defect features with a preset standard defect library. This standard library stores standard feature vectors for various types of defects (such as different levels of insulator damage, hardware corrosion, conductor damage, etc.). Through comparison, the system can automatically identify the defect type and classify the defect according to the degree of matching and feature parameters (such as area and depth).

[0059] The classification results of construction process defects can include Level 1 defects, Level 2 defects, and Level 3 defects; Level 1 defects can be emergency defects that directly threaten the safety of the line (such as insulator breakage). Level 2 defects can be serious defects that require treatment within a specified period (such as the area of ​​wire corrosion > 50%). Level 3 defects can be general defects that require recording and observation (such as insulator contamination).

[0060] At the same time, the system will automatically verify the equipment model identified by OCR with the equipment model in the project ledger in step 101 to ensure that the equipment installed on site is consistent with the design requirements, prevent incorrect installation or omission, mark the inconsistencies and generate a difference report.

[0061] Step 106: Based on the grading results of construction process defects and the verification results of equipment models, generate and output an acceptance report.

[0062] Finally, the system can summarize all defect information (such as type, location, level, and photographic evidence) and equipment verification results (such as consistency / inconsistency) and automatically generate a standardized and detailed acceptance report. The acceptance report can be transmitted back to the ground management platform in real time via wireless network for acceptance personnel to review, archive, and use as the basis for subsequent rectification work.

[0063] In summary, according to the power distribution network drone acceptance method provided in this application, compared with the existing technology, this application can generate an initial drone inspection route based on the engineering ledger and geographic information data of the power distribution network overhead lines; based on the initial drone inspection route, visual fusion tracking technology is adopted to integrate multi-dimensional real-time visual information and combine it with a target centering algorithm to achieve real-time identification and continuous tracking of power distribution equipment; during the tracking process, combined with multi-dimensional real-time visual information, the drone's flight altitude and heading are dynamically adjusted through an AI autonomous obstacle avoidance model to collect images of the power distribution facilities; the images of the power distribution facilities are processed by an airborne edge computing intelligent terminal, and image recognition technology is applied to extract construction process defect features and equipment nameplate identification features from the processed images; the construction process defect features are compared with a preset standard library to identify and classify the construction process defects of the power distribution equipment, and the equipment model of the power distribution equipment is identified and obtained based on the extracted equipment nameplate identification features, and the model is verified with the equipment model in the engineering ledger; based on the classification results of the construction process defects and the verification results of the equipment model, an acceptance report is generated and output. By adopting the above technical solution, this application automatically generates initial inspection routes covering all power distribution equipment based on engineering ledgers and geographic information data, eliminating the need for manual on-site surveys and reducing fieldwork workload by more than 50%. From route planning and construction inspection to acceptance report generation, no manual intervention is required throughout the entire process. The drone automatically performs data collection, analysis, and decision-making tasks, with data transmitted back to the cloud in real time, forming a closed loop of "collection-processing-decision-feedback," reducing the acceptance cycle and improving efficiency.

[0064] This application utilizes visual fusion tracking technology, combined with a target centering algorithm, to ensure that the drone continuously locks onto power distribution equipment in complex environments (such as strong winds or obstructions), avoiding repeated inspections due to target loss. Combined with an AI-powered autonomous obstacle avoidance model, the drone can safely and flexibly approach the equipment, avoiding image blurring caused by excessive distance or angular deviations. This ensures data quality from the source, reducing misjudgments and missed detections due to poor image quality, thus meeting the needs of rapid and accurate acceptance testing for large-scale power distribution network projects.

[0065] Based on the above Figure 1 The specific implementation of the method shown in this embodiment provides a power distribution network drone acceptance device, such as... Figure 2 As shown, the device includes: a generation module 31, a tracking module 32, a acquisition module 33, an extraction module 34, an identification module 35, and an output module 36; The generation module 31 is used to generate the initial UAV inspection route based on the engineering ledger and geographic information data of the overhead power distribution line. The tracking module 32 is used to achieve real-time identification and continuous tracking of power distribution equipment based on the initial UAV inspection route, using visual fusion tracking technology, fusing multi-dimensional real-time visual information, and combining a target centering algorithm. The acquisition module 33 is used to dynamically adjust the flight altitude and heading of the drone by combining the multi-dimensional real-time visual information and using an AI autonomous obstacle avoidance model during the tracking process, so as to acquire images of the power distribution equipment's distribution network facilities. Extraction module 34 is used to process the image of the power distribution facility through an airborne edge computing smart terminal, and to extract construction process defect features and equipment nameplate identification features from the processed image of the power distribution facility using image recognition technology. The identification module 35 is used to compare the construction process defect features with a preset standard library to identify and classify the construction process defects of the power distribution equipment. At the same time, based on the extracted equipment nameplate identification features, the module identifies and obtains the equipment model of the power distribution equipment and performs model verification with the equipment model in the project ledger. Output module 36 is used to generate and output an acceptance report based on the grading results of the construction process defects and the verification results of the equipment model.

[0066] In specific application scenarios, the tracking module 32 can be used to collect images of the power distribution area of ​​the power distribution equipment through the visible light camera of the drone based on the initial drone inspection route. The relative distance and angle data between the UAV and the current visual target are collected synchronously by the visual ranging module of the UAV, and the power distribution area image is fused with the relative distance and angle data to generate fused multi-dimensional real-time visual information. Based on the fused multi-dimensional real-time visual information, a pre-trained target detection model is invoked to identify the pixel region of the power distribution equipment in the power distribution area image. The pixel coordinates of the power distribution equipment are locked and recorded by the target box, so as to determine the physical center coordinates of the power distribution area image based on the pixel coordinates. The target centering algorithm calculates the coordinate offset between the center coordinates of the target box and the physical center coordinates, and dynamically adjusts the flight attitude of the UAV based on the coordinate offset to keep the target box centered in the image, thereby achieving real-time identification and continuous tracking of the power distribution equipment.

[0067] In specific application scenarios, the acquisition module 33 can be used to call the obstacle recognition model to perform data analysis on the multi-dimensional real-time visual information and identify obstacles in the environment, including obstacle type and obstacle size; The actual distance between the obstacle and the drone is calculated using visual ranging data and lidar data, as well as the vertical height difference between the top of the obstacle and the power line. Based on the actual distance value and / or the vertical height difference, the risk level of the obstacle is determined, and based on the risk level, an obstacle avoidance command is output. The drone's flight altitude and heading are dynamically adjusted according to the obstacle avoidance command through an AI autonomous obstacle avoidance model, and images of the power distribution equipment's distribution network facilities are collected.

[0068] In specific application scenarios, the data acquisition module 33 can be used to determine the risk level of the obstacle as high risk level if the actual distance value and / or the vertical height difference is less than the minimum preset threshold, or if the obstacle intrudes into the inspection route range. If the actual distance value and / or the vertical height difference is greater than or equal to the minimum preset threshold, and the actual distance value and / or the vertical height difference is less than the maximum preset threshold, and the obstacle does not intrude into the inspection route range, the risk level of the obstacle is determined to be medium risk level. If the actual distance value and / or the vertical height difference is greater than or equal to the maximum preset threshold, the risk level of the obstacle is determined to be low risk level.

[0069] In specific application scenarios, the extraction module 34 can be used to segment the power distribution equipment image in the power distribution facility image using a lightweight semantic segmentation model, locate the defect region in the power distribution equipment image by threshold filtering, and extract the geometric features and surface texture features of the defect region to obtain the construction process defect features. The geometric features include the defect region area calculated based on the number of pixels in the defect region and the image pixel resolution; and / or, the roundness calculated based on the defect region area; the surface texture features include contrast and / or correlation calculated based on the gray-level co-occurrence matrix.

[0070] In specific application scenarios, the extraction module 34 can be used to detect the nameplate area in the image of the power distribution facility; after filtering out false targets in the nameplate area by contour rectangle, the geometric features and surface texture features of the characters in the nameplate area are extracted, and optical character recognition technology is used to recognize the characters based on the geometric features and surface texture features, and the recognition result is used as the identification feature of the equipment nameplate.

[0071] It should be noted that other corresponding descriptions of the functional units involved in the power distribution network drone acceptance device provided in this embodiment can be found in [reference]. Figure 1 The corresponding descriptions in [the document] will not be repeated here.

[0072] Based on the above, Figure 1Accordingly, this embodiment also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method. Figure 1 The method shown.

[0073] Based on this understanding, the technical solution of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause a computer device (such as personal computer, server, or network device, etc.) to execute the methods of various implementation scenarios of this application.

[0074] Based on the above, Figure 1 The method shown, and Figure 2 To achieve the above objectives, the present application also provides an electronic device, comprising a storage medium and a processor; the storage medium for storing a computer program; and the processor for executing the computer program to implement the above-described virtual device embodiments. Figure 1 The method shown.

[0075] Optionally, the aforementioned physical devices may also include a user interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, a Wi-Fi module, etc. The user interface may include a display screen, input units such as a keyboard, etc., and optional user interfaces may also include USB interfaces, card reader interfaces, etc. The network interface may optionally include standard wired interfaces, wireless interfaces (such as Wi-Fi interfaces), etc.

[0076] Those skilled in the art will understand that the physical device structure provided in this embodiment does not constitute a limitation on the physical device, and may include more or fewer components, or combine certain components, or have different component arrangements.

[0077] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the aforementioned physical device, supporting the operation of the power distribution drone acceptance program and other software and / or programs. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software in the power distribution drone acceptance physical device.

[0078] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platform, or it can be implemented by hardware. By applying the solution of this embodiment, compared with the existing technology, this application can generate an initial UAV inspection route based on the engineering ledger and geographic information data of the overhead power distribution line; based on the initial UAV inspection route, visual fusion tracking technology is adopted to integrate multi-dimensional real-time visual information and combine it with a target centering algorithm to achieve real-time identification and continuous tracking of power distribution equipment; during the tracking process, combined with multi-dimensional real-time visual information, the UAV's flight altitude and heading are dynamically adjusted through an AI autonomous obstacle avoidance model to collect images of the power distribution equipment's power distribution facilities; the images of the power distribution facilities are processed by an airborne edge computing intelligent terminal, and image recognition technology is applied to extract construction process defect features and equipment nameplate identification features from the processed images of the power distribution facilities; the construction process defect features are compared with a preset standard library to identify and classify the construction process defects of the power distribution equipment, and the equipment model of the power distribution equipment is identified and obtained based on the extracted equipment nameplate identification features, and the model is verified with the equipment model in the engineering ledger; based on the classification results of the construction process defects and the verification results of the equipment model, an acceptance report is generated and output. By adopting the above technical solution, this application automatically generates initial inspection routes covering all power distribution equipment based on engineering ledgers and geographic information data, eliminating the need for manual on-site surveys and reducing fieldwork workload by more than 50%. From route planning and construction inspection to acceptance report generation, no manual intervention is required throughout the entire process. The drone automatically performs data collection, analysis, and decision-making tasks, with data transmitted back to the cloud in real time, forming a closed loop of "collection-processing-decision-feedback," reducing the acceptance cycle and improving efficiency.

[0079] This application utilizes visual fusion tracking technology, combined with a target centering algorithm, to ensure that the drone continuously locks onto power distribution equipment in complex environments (such as strong winds or obstructions), avoiding repeated inspections due to target loss. Combined with an AI-powered autonomous obstacle avoidance model, the drone can safely and flexibly approach the equipment, avoiding image blurring caused by excessive distance or angular deviations. This ensures data quality from the source, reducing misjudgments and missed detections due to poor image quality, thus meeting the needs of rapid and accurate acceptance testing for large-scale power distribution network projects.

[0080] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the term "comprising" or any other variations thereof is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0081] The above are merely specific embodiments of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to these embodiments, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for acceptance of a network configuration unmanned aerial vehicle based on visual fusion tracking and image recognition, characterized in that, The method comprises: generating an initial unmanned aerial vehicle inspection route based on the engineering account and geographic information data of the power distribution overhead line; based on the initial unmanned aerial vehicle inspection route, using visual fusion tracking technology to fuse multi-dimensional real-time visual information, and combining a target centering algorithm to realize real-time identification and continuous tracking of the power distribution equipment; in the tracking process, combining the multi-dimensional real-time visual information, dynamically adjusting the flight height and heading of the unmanned aerial vehicle through an AI autonomous obstacle avoidance model, to collect power distribution facility images of the power distribution equipment; processing the power distribution facility images through an on-board edge computing intelligent terminal, and applying image recognition technology to extract construction process defect features and equipment nameplate identification features from the processed power distribution facility images; comparing the construction process defect features with a preset standard library to identify and grade the construction process defects of the power distribution equipment, and identifying the equipment model of the power distribution equipment according to the extracted equipment nameplate identification features, and verifying the equipment model with the equipment model in the engineering account; generating and outputting an acceptance report according to the grading results of the construction process defects and the verification results of the equipment model.

2. The method of claim 1, wherein, based on the initial unmanned aerial vehicle inspection route, using visual fusion tracking technology to fuse multi-dimensional real-time visual information, and combining a target centering algorithm to realize real-time identification and continuous tracking of the power distribution equipment, comprising: based on the initial unmanned aerial vehicle inspection route, collecting power distribution area images of the power distribution equipment through a visible light camera of the unmanned aerial vehicle; synchronously collecting relative distance and angle data between the unmanned aerial vehicle and the current visual target through a visual ranging module of the unmanned aerial vehicle, and fusing the power distribution area images with the relative distance and the angle data to generate fused multi-dimensional real-time visual information; based on the fused multi-dimensional real-time visual information, calling a pre-trained target detection model to identify the pixel area of the power distribution equipment in the power distribution area images, locking and recording the pixel coordinates of the power distribution equipment through a target box, to determine the physical center coordinates of the power distribution area images according to the pixel coordinates; calculating the coordinate offset between the center coordinates of the target box and the physical center coordinates through the target centering algorithm, and dynamically adjusting the flight attitude of the unmanned aerial vehicle according to the coordinate offset, to keep the target box centered in the image, realizing real-time identification and continuous tracking of the power distribution equipment.

3. The method of claim 1, wherein, in the tracking process, combining the multi-dimensional real-time visual information, dynamically adjusting the flight height and heading of the unmanned aerial vehicle through an AI autonomous obstacle avoidance model, to collect power distribution facility images of the power distribution equipment, comprising: calling an obstacle recognition model to analyze the multi-dimensional real-time visual information, identifying obstacles in the environment, the obstacles including obstacle type and obstacle size; calculating the actual distance value between the obstacles and the unmanned aerial vehicle through visual ranging data and laser radar data, and calculating the vertical height difference value between the top of the obstacle and the power distribution line; According to the actual distance value and / or the vertical height difference value, a risk level of the obstacle is determined, and according to the risk level, a obstacle-avoiding instruction is outputted, so that the flight height and the heading of the unmanned aerial vehicle are dynamically adjusted through an AI autonomous obstacle-avoiding model according to the obstacle-avoiding instruction, and the power distribution equipment image is collected.

4. The method of claim 3, wherein, The risk level of the obstacle is determined according to the actual distance value and / or the vertical height difference value, including: If the actual distance value and / or the vertical height difference is less than a minimum preset threshold, or the obstacle invades the inspection route range, it is determined that the risk level of the obstacle is a high risk level; If the actual distance value and / or the vertical height difference is greater than or equal to the minimum preset threshold, and the actual distance value and / or the vertical height difference is less than a maximum preset threshold, and the obstacle does not invade the inspection route range, it is determined that the risk level of the obstacle is a medium risk level; If the actual distance value and / or the vertical height difference is greater than or equal to the maximum preset threshold, it is determined that the risk level of the obstacle is a low risk level.

5. The method of claim 1, wherein, The construction process defect features are extracted from the processed power distribution equipment image, including: The power distribution equipment image in the power distribution equipment image is segmented through a lightweight semantic segmentation model, and the defect area in the power distribution equipment image is located through threshold screening; The geometric features and surface texture features of the defect area are extracted to obtain the construction process defect features.

6. The method of claim 5, wherein, The geometric features include a defect area area calculated according to the number of pixels in the defect area and the image pixel resolution; And / or, a circularity calculated according to the defect area area; and the surface texture features include a contrast and / or a correlation calculated based on a gray level co-occurrence matrix.

7. The method of claim 1, wherein, The equipment nameplate identification features are extracted from the processed power distribution equipment image, including: The nameplate area in the power distribution equipment image is detected; After excluding false targets in the nameplate area through contour rectangularity screening, the geometric features and surface texture features of characters in the nameplate area are extracted, and an optical character recognition technology is used to perform character recognition on the characters according to the geometric features and the surface texture features, and the recognition result is taken as the equipment nameplate identification features.

8. A power distribution network drone acceptance device, characterized in that, Including: A generation module is configured to generate an initial unmanned aerial vehicle inspection route based on engineering account and geographic information data of a power distribution overhead line; A tracking module is configured to, based on the initial unmanned aerial vehicle inspection route, use a visual fusion tracking technology to fuse multi-dimensional real-time visual information, and combine a target centering algorithm to realize real-time identification and continuous tracking of power distribution equipment; An acquisition module is configured to, in the tracking process, combine the multi-dimensional real-time visual information, dynamically adjust the flight height and the heading of the unmanned aerial vehicle through an AI autonomous obstacle-avoiding model, and collect a power distribution equipment image; An extraction module is configured to process the power distribution equipment image through an on-board edge computing intelligent terminal, and extract construction process defect features and equipment nameplate identification features from the processed power distribution equipment image by using an image recognition technology. The identification module is configured to compare the construction process defect feature with a preset standard library, to identify and grade the construction process defect of the power distribution equipment, and to identify and acquire a device model of the power distribution equipment according to the extracted device nameplate identification feature, and to perform model verification with a device model in the engineering account; The output module is configured to generate and output an acceptance report according to the grading result of the construction process defect and the verification result of the device model.

9. An electronic device comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, The processor executes the computer program to implement the method in any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the method in any one of claims 1 to 7.

Images