Unmanned aerial vehicle photovoltaic inspection positioning method and system based on extended Kalman filtering

Abstract

The invention relates to the technical field of photovoltaic inspection, in particular to an unmanned aerial vehicle photovoltaic inspection positioning method and system based on extended Kalman filtering, and the method comprises the steps: obtaining the GPS positioning data, attitude data and photovoltaic panel image data of an unmanned aerial vehicle in real time based on a sensor carried by the unmanned aerial vehicle; according to the image data of the photovoltaic panel, fault features of the fault photovoltaic panel are identified, and pixel coordinates of the fault photovoltaic panel are extracted; converting the pixel coordinates of the fault photovoltaic panel into initial position information under a geographic coordinate system; constructing an extended Kalman filtering model, inputting the positioning data, attitude data and initial position information of the unmanned aerial vehicle into the extended Kalman filtering model, and establishing a state vector and an observation vector; and iteratively calculating an optimal estimation value through an extended Kalman filtering algorithm according to the state vector, and outputting geographic coordinates of the fault photovoltaic panel. According to the invention, the method achieves the quick and precise positioning of the fault photovoltaic panel, improves the operation and maintenance efficiency of a photovoltaic power station, and reduces the operation and maintenance cost.

Description

Technical Field

[0001] This invention relates to the technical field of photovoltaic inspection, and in particular to a UAV photovoltaic inspection positioning method and system based on extended Kalman filtering. Background Technology

[0002] With the increasing global demand for clean energy, photovoltaic (PV) power plants, as an efficient and environmentally friendly energy production method, have been constructed and operated on a large scale. However, the efficient operation of PV power plants relies on regular inspection and maintenance of the photovoltaic panels. During long-term operation, PV panels may experience various faults, such as hot spot effects, cracks, shading, and aging. These faults can affect the power generation efficiency of the panels and even lead to equipment damage. Therefore, timely and accurate detection and handling of these faults are crucial to ensuring the stable operation of PV power plants.

[0003] Currently, the application of drone technology in the field of photovoltaic inspection has gradually become widespread. Drones have advantages such as high flexibility, wide coverage, and the ability to quickly reach designated locations, effectively compensating for the shortcomings of manual inspection. By equipping themselves with sensors such as high-definition cameras and infrared thermal imagers, drones can quickly collect image and video data of photovoltaic panels, providing a foundation for subsequent analysis and diagnosis.

[0004] However, traditional UAV positioning methods primarily rely on the Global Positioning System (GPS). While GPS provides relatively accurate location information, its accuracy is significantly affected in complex environments (such as mountainous areas or areas obstructed by buildings). Furthermore, GPS positioning accuracy typically only reaches the meter level, which is insufficient to meet the precise location requirements of faulty photovoltaic panels during photovoltaic inspections. For example, in large-scale photovoltaic power plants, relying solely on GPS positioning may not accurately distinguish adjacent photovoltaic panels, leading to inaccurate fault location and increasing the search time for maintenance personnel. Therefore, there is room for improvement. Summary of the Invention

[0005] To achieve rapid and accurate location of faulty photovoltaic panels, improve the operation and maintenance efficiency of photovoltaic power plants, and reduce operation and maintenance costs, this application provides a method and system for unmanned aerial vehicle (UAV) photovoltaic inspection and location based on extended Kalman filtering.

[0006] Firstly, the above-mentioned inventive objective of this application is achieved through the following technical solution:

[0007] A method for UAV photovoltaic inspection and positioning based on extended Kalman filtering, comprising the following steps:

[0008] The drone's GPS positioning data, attitude data, and photovoltaic panel image data are acquired in real time using sensors mounted on the drone.

[0009] Based on the image data of the photovoltaic panel, identify the fault characteristics of the faulty photovoltaic panel and extract its pixel coordinates;

[0010] Convert the pixel coordinates of the faulty photovoltaic panel into initial location information in the geographic coordinate system;

[0011] An extended Kalman filter model is constructed by inputting the UAV's positioning data, attitude data, and initial position information into the extended Kalman filter model to establish state vectors and observation vectors.

[0012] The optimal estimate is calculated iteratively using the extended Kalman filter algorithm based on the state vector, and the geographical coordinates of the faulty photovoltaic panel are output.

[0013] By adopting the above technical solution, during the inspection of photovoltaic panels using drones, the multi-source sensors carried by the drone, including GPS, IMU, and vision sensors, are used to acquire GPS positioning data, attitude data, and photovoltaic panel image data in real time. By fusing the absolute position information of GPS, the high-frequency attitude data of IMU, and the image acquisition capabilities of the vision sensor, the signal obstruction problem caused by relying solely on GPS is effectively overcome. In mountainous areas or densely built-up areas, this ensures the comprehensiveness and reliability of the data source, laying a solid foundation for subsequent processing. In the stage of identifying the fault characteristics of the photovoltaic panel and extracting its pixel coordinates, the vision sensor is used to analyze the image data in real time, quickly detecting fault characteristics such as hot spot effects, cracks, or occlusion. The pixel coordinates are accurately extracted through edge detection algorithms, reducing the risk of subjective misjudgment in traditional manual inspection and improving the accuracy and efficiency of fault identification. When converting the pixel coordinates of the faulty photovoltaic panel into the initial position information in the geographic coordinate system, a transformation matrix is ​​established by combining the real-time altitude, pitch angle, and yaw angle data of the drone, and calibrated with GPS data to generate a highly reliable initial position estimate. This solves the positioning error caused by inaccurate coordinate transformation in existing technologies. To address the issue of poor performance, high-quality input was provided for subsequent filtering. In the stage of constructing the Extended Kalman Filter (EKF) model, the UAV's positioning data, attitude data, and initial position information obtained in the previous steps were used as input to establish a state vector containing the UAV's three-dimensional position, velocity, attitude angle, and coordinates of the faulty photovoltaic panel, as well as an observation vector composed of multiple source sensors. This design fully utilizes the advantages of EKF in handling nonlinear systems, achieving deep fusion of multi-source data through state transition equations and observation equations. This effectively compensates for data noise in complex environments (such as light interference or GPS signal loss), significantly enhancing robustness and adaptability. In the process of iteratively calculating the optimal estimate and outputting the geographical coordinates of the faulty photovoltaic panel based on the state vector using the EKF algorithm, the EKF algorithm achieves centimeter-level accuracy in fault location output through state vector prediction in the prediction stage, linearization approximation of the Jacobian matrix, and Kalman gain calculation in the update stage. This ensures that fault information can be fed back to maintenance personnel in real time, not only significantly improving positioning accuracy to the centimeter level but also enhancing stability in complex environments such as shaded or mountainous areas. Ultimately, this reduces the operation and maintenance costs and fault response time of the photovoltaic power station, and improves power generation efficiency and system reliability.

[0014] In a preferred embodiment, this application can be further configured such that: the step of identifying the fault features of a faulty photovoltaic panel based on its image data and extracting its pixel coordinates specifically includes:

[0015] The image data of the photovoltaic panel is preprocessed to eliminate light interference and background noise, resulting in preprocessed photovoltaic panel image data.

[0016] Temperature threshold segmentation is performed on the preprocessed photovoltaic panel image data to extract abnormal temperature areas and locate fault points.

[0017] The contour boundaries of the preprocessed photovoltaic panel image data are extracted based on the edge detection algorithm, and the pixel coordinate system of the photovoltaic panel image data is generated based on the contour boundaries.

[0018] Calculate the pixel coordinates of the fault point center in the preprocessed photovoltaic panel image data and associate them with timestamp information.

[0019] By adopting the above technical solutions, image quality is significantly improved through preprocessing of image data (eliminating illumination interference and background noise), solving the visual interference problem in complex lighting environments (such as mountainous or shaded areas), reducing the risk of misjudgment due to shadows or reflections, and ensuring the reliability of subsequent analysis. Temperature threshold segmentation is used to quickly extract abnormal temperature areas (such as hot spots or localized overheating), directly locating fault points and efficiently identifying common photovoltaic panel faults (such as hot spot effects or aging). An edge detection algorithm is used to extract image contour boundaries and generate a pixel coordinate system, accurately delineating the photovoltaic panel's geometry and avoiding the problems of blurred or discontinuous boundaries in traditional methods. This provides a high-precision foundation for coordinate transformation, supporting centimeter-level positioning requirements. The center pixel coordinates of the fault point are calculated and associated with timestamp information, achieving synchronous recording of space and time, reducing data processing delays, and ensuring that fault location information is fed back to the positioning system in real time.

[0020] In a preferred embodiment, this application can be further configured such that converting the pixel coordinates of the faulty photovoltaic panel into initial location information in a geographic coordinate system specifically includes:

[0021] Based on the intrinsic parameter matrix and distortion coefficients of the drone's camera, the pixel coordinates are subjected to distortion correction.

[0022] By combining the real-time altitude, pitch angle, and yaw angle data of the UAV, a transformation matrix from pixel coordinates to geographic coordinates is established;

[0023] The distortion-free pixel coordinates are mapped to the geographic coordinate system using a perspective transformation model;

[0024] The conversion results are calibrated by integrating GPS positioning data to generate initial location information.

[0025] By adopting the above technical solution, pixel coordinates are distorted (using the camera intrinsic matrix and distortion coefficients of the UAV to correct lens distortion), effectively eliminating barrel or pincushion distortion during image acquisition and solving coordinate offset problems caused by camera lens defects in complex environments. For example, it reduces pixel errors in mountainous or strong light scenes, providing high-fidelity input for subsequent conversion. When establishing the transformation matrix from pixel coordinates to geographic coordinates, a mathematical model based on three-dimensional spatial geometry is constructed by combining the real-time altitude, pitch angle, and yaw angle data of the UAV. This accurately handles perspective distortion when the UAV is flying at an angle, ensuring consistency of coordinate mapping under different perspectives and avoiding scale distortion caused by attitude changes in traditional methods. The perspective transformation model maps the distorted pixel coordinates to the geographic coordinate system, achieving a one-to-one correspondence between pixel points and ground positions. Affine transformation or projection algorithms are used to efficiently process large-scale photovoltaic arrays. GPS positioning data is fused to calibrate the conversion results, compensate for cumulative errors caused by environmental interference (such as signal blockage), and generate robust initial position information, providing high-precision observation input for subsequent extended Kalman filtering.

[0026] In a preferred embodiment, this application can be further configured as follows: the construction of the extended Kalman filter model involves inputting the UAV's positioning data, attitude data, and initial position information into the extended Kalman filter model to establish a state vector and an observation vector, specifically including:

[0027] An observation vector input is generated based on the initial position information of the faulty photovoltaic panel and the GPS positioning and attitude data of the UAV;

[0028] The three-dimensional position, velocity, attitude angle of the UAV, and the geographic coordinates of the faulty photovoltaic panel to be located are defined as a joint state vector;

[0029] A nonlinear state transition equation is constructed based on the attitude data of the UAV to describe the evolution of the joint state vector over time.

[0030] The joint state vector is mapped to the initial position information of the multi-source sensors and the faulty photovoltaic panel carried by the UAV to form the observation equation.

[0031] By adopting the above technical solution, the initial position information of the faulty photovoltaic panel is integrated with the GPS positioning data and attitude data of the UAV to generate the observation vector input. This fully utilizes the complementarity of multi-source data (such as the absolute position accuracy of GPS, the high-frequency attitude update of the IMU, and the local accuracy of the initial visual positioning), effectively overcoming the data loss problem of a single sensor in obstructed or interfered environments, and improving system robustness. The UAV's three-dimensional position, velocity, attitude angle, and the geographical coordinates of the faulty photovoltaic panel are defined as a joint state vector, breaking through the limitation of traditional positioning methods that only focus on the UAV's own state. This achieves cooperative state estimation of the UAV and the faulty photovoltaic panel, significantly improving fault location accuracy to the centimeter level. Based on the UAV's attitude data, a nonlinear state transition equation is constructed to accurately describe the UAV's kinematic laws and the position of the faulty photovoltaic panel over time. The evolutionary relationship is used to approximate the nonlinear system through Taylor series linearization, effectively suppressing state estimation drift in complex environments and ensuring the stability of photovoltaic panel fault location. The observation equation is designed to map the joint state vector to multi-source sensor data and initial position information, forming a closed-loop feedback mechanism. The state estimation weight is dynamically adjusted through Kalman gain to correct the positioning deviation in real time. At the same time, the efficient data compression and parallel computing are combined to optimize the processing flow, reduce positioning latency, and provide maintenance personnel with real-time and accurate fault location information. The extended Kalman filter model maintains centimeter-level positioning accuracy in complex scenarios such as mountainous areas with obstruction by jointly defining the state vector (covering both UAV and photovoltaic panel targets) and multi-source fusion of the observation equation (integrating GPS / IMU / visual data). This significantly improves the operation and maintenance efficiency of photovoltaic power plants and reduces maintenance costs.

[0032] In a preferred embodiment, this application can be further configured such that: the step of iteratively calculating the optimal estimate based on the state vector using an extended Kalman filter algorithm and outputting the geographical coordinates of the faulty photovoltaic panel specifically includes:

[0033] Predict the state vector and covariance matrix at the next time step based on the nonlinear state transition equation;

[0034] The nonlinear state transition equation is linearized and approximated to solve for the state transition matrix and the observation matrix.

[0035] Calculate the Kalman gain by combining sensor observations, and update the state vector and covariance matrix;

[0036] Repeat the above steps iteratively until the state estimation error converges to a centimeter-level threshold.

[0037] By adopting the above technical solution, the state vector and covariance matrix of the next moment are predicted based on the nonlinear state transition equation. The state evolution trend is predicted by using the UAV kinematic model (such as the correlation between the rate of change of attitude angle and velocity), effectively compensating for the positioning interruption problem caused by GPS signal loss or obstruction. It maintains prediction stability even in complex environments such as strong winds in mountainous areas. By linearizing the nonlinear state transition equation (using Taylor series expansion to solve the Jacobian matrix), nonlinear variables such as UAV attitude angle and faulty photovoltaic panel position are transformed into a computable linear system, overcoming the limitations of traditional linear filtering algorithms in photovoltaic panel tilt installation scenarios, and improving the positioning accuracy to the centimeter level. Then, combined with sensor observations (GPS / IMU / initial data), the system can be further improved. The system calculates the Kalman gain in real time using initial position information, dynamically adjusts the state estimation weights (e.g., increasing the weight of visual initial position data when GPS signals are weak), and achieves closed-loop feedback calibration of multi-source data by updating the state vector and covariance matrix. This significantly suppresses data noise interference (such as IMU drift or image coordinate transformation deviation), compressing the fluctuation range of positioning results to within the centimeter level. The above steps are iteratively executed until the state estimation error converges to the centimeter-level threshold, ensuring that the output results meet the requirements of photovoltaic panel-level precise positioning. The overall technical solution, through a closed-loop iterative mechanism of prediction-linearization-update-convergence, significantly enhances the system's adaptability in occlusion, strong light, and dynamic flight scenarios while ensuring centimeter-level positioning accuracy, and reduces fault location delay.

[0038] In a preferred embodiment, this application can be further configured such that the linearization approximation of the nonlinear state transition equation specifically includes:

[0039] The Jacobian matrix is ​​solved by using the first-order partial derivatives of the state transition equation through Taylor series expansion;

[0040] Correlate the rate of change of the coordinates of the faulty photovoltaic panel with the rate of change of the drone's attitude angle;

[0041] Compensate for state prediction errors when GPS signals are lost by using attitude data;

[0042] Quantifying sensor data uncertainty based on process noise covariance matrix.

[0043] By adopting the above technical solution, the Jacobian matrix is ​​generated by solving the first-order partial derivatives of the state transition equation through Taylor series expansion. This transforms nonlinear variables such as the UAV attitude angle and the position of the faulty photovoltaic panel into a computable linear system (e.g., relating the pitch angle change rate to the photovoltaic panel coordinate derivative). This overcomes the limitations of traditional algorithms in scenarios with tilted photovoltaic panel installations, improves model computational efficiency, and dynamically correlates the faulty photovoltaic panel coordinate change rate with the UAV attitude angle change rate (e.g., correcting coordinate offsets using the roll rate obtained in real-time from the IMU). This accurately describes the impact of complex maneuvering flight (e.g., sharp turns during mountain inspections) on positioning, solving the position drift problem in dynamic scenarios. Errors caused by attitude interference are reduced by compensating for state prediction errors when GPS signals are lost using attitude data (e.g., using IMU angular velocity integrals to replace GPS positioning in obscured areas). The prediction mechanism of extended Kalman filtering is used to maintain system stability, ensuring that positioning accuracy remains at the centimeter level even when GPS is interrupted for a certain period of time. Sensor data uncertainty is quantified based on the process noise covariance matrix (e.g., assigning differentiated noise weights to GPS / IMU / visual data). The multi-source data fusion ratio is adaptively adjusted through Kalman gain to suppress sensor noise interference (e.g., IMU zero bias or residual image distortion errors), thus compressing the fluctuation range of positioning results to within the centimeter level.

[0044] In a preferred embodiment, this application can be further configured such that, after the optimal estimate is iteratively calculated using the extended Kalman filter algorithm based on the state vector and the geographical coordinates of the faulty photovoltaic panel are output, the UAV photovoltaic inspection and positioning method based on extended Kalman filter further includes:

[0045] The geographic coordinates output by the extended Kalman filter are mapped to the photovoltaic panel array number to generate an inspection report containing the fault location, fault type, and confidence level.

[0046] The report is pushed to the operation and maintenance terminal in real time via wireless network, and the coordinates of the fault point are dynamically marked in the drone flight path planning.

[0047] By adopting the above technical solution, the problem of the disconnect between traditional positioning results and physical location is solved by accurately mapping geographical coordinates to photovoltaic panel array numbers (such as raster matching based on the power plant GIS database). This allows maintenance personnel to directly locate the faulty panel number and generate a multi-dimensional inspection report containing the fault location, fault type (such as hot spot / crack / shading), and confidence level (based on EKF covariance matrix calculation). By integrating infrared thermal imaging data and visual analysis results, intelligent diagnosis of fault nature is achieved. The report is pushed to the maintenance terminal (such as mobile device or central control system) in real time via 4G / 5G wireless network. The fault point coordinates are dynamically marked in the drone flight path planning (such as automatic obstacle avoidance or close-range re-inspection). The optimal inspection path is constructed by combining the real-time updated digital map.

[0048] Secondly, the above-mentioned inventive objective of this application is achieved through the following technical solutions:

[0049] A UAV photovoltaic inspection and positioning system based on extended Kalman filtering, the system comprising:

[0050] The data acquisition module is used to acquire GPS positioning data, attitude data, and photovoltaic panel image data of the UAV in real time based on the sensors carried by the UAV.

[0051] A photovoltaic panel fault identification module is used to identify the fault characteristics of a faulty photovoltaic panel based on the image data of the photovoltaic panel and extract its pixel coordinates;

[0052] The coordinate transformation module is used to convert the pixel coordinates of the faulty photovoltaic panel into initial location information in the geographic coordinate system;

[0053] The model building module is used to build an extended Kalman filter model. The positioning data, attitude data and initial position information of the UAV are input into the extended Kalman filter model to establish the state vector and observation vector.

[0054] The photovoltaic panel fault coordinate output module is used to iteratively calculate the optimal estimate based on the state vector using the extended Kalman filter algorithm, and output the geographical coordinates of the faulty photovoltaic panel.

[0055] By adopting the above technical solution, during the inspection of photovoltaic panels using drones, the multi-source sensors carried by the drone, including GPS, IMU, and vision sensors, are used to acquire GPS positioning data, attitude data, and photovoltaic panel image data in real time. By fusing the absolute position information of GPS, the high-frequency attitude data of IMU, and the image acquisition capabilities of the vision sensor, the signal obstruction problem caused by relying solely on GPS is effectively overcome. In mountainous areas or densely built-up areas, this ensures the comprehensiveness and reliability of the data source, laying a solid foundation for subsequent processing. In the stage of identifying the fault characteristics of the photovoltaic panel and extracting its pixel coordinates, the vision sensor is used to analyze the image data in real time, quickly detecting fault characteristics such as hot spot effects, cracks, or occlusion. The pixel coordinates are accurately extracted through edge detection algorithms, reducing the risk of subjective misjudgment in traditional manual inspection and improving the accuracy and efficiency of fault identification. When converting the pixel coordinates of the faulty photovoltaic panel into the initial position information in the geographic coordinate system, a transformation matrix is ​​established by combining the real-time altitude, pitch angle, and yaw angle data of the drone, and calibrated with GPS data to generate a highly reliable initial position estimate. This solves the positioning error caused by inaccurate coordinate transformation in existing technologies. To address the issue of poor performance, high-quality input was provided for subsequent filtering. In the stage of constructing the Extended Kalman Filter (EKF) model, the UAV's positioning data, attitude data, and initial position information obtained in the previous steps were used as input to establish a state vector containing the UAV's three-dimensional position, velocity, attitude angle, and coordinates of the faulty photovoltaic panel, as well as an observation vector composed of multiple source sensors. This design fully utilizes the advantages of EKF in handling nonlinear systems, achieving deep fusion of multi-source data through state transition equations and observation equations. This effectively compensates for data noise in complex environments (such as light interference or GPS signal loss), significantly enhancing robustness and adaptability. In the process of iteratively calculating the optimal estimate and outputting the geographical coordinates of the faulty photovoltaic panel based on the state vector using the EKF algorithm, the EKF algorithm achieves centimeter-level accuracy in fault location output through state vector prediction in the prediction stage, linearization approximation of the Jacobian matrix, and Kalman gain calculation in the update stage. This ensures that fault information can be fed back to maintenance personnel in real time, not only significantly improving positioning accuracy to the centimeter level but also enhancing stability in complex environments such as shaded or mountainous areas. Ultimately, this reduces the operation and maintenance costs and fault response time of the photovoltaic power station, and improves power generation efficiency and system reliability.

[0056] Thirdly, the above-mentioned objectives of this application are achieved through the following technical solutions:

[0057] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the above-described UAV photovoltaic inspection and positioning method based on extended Kalman filtering.

[0058] Fourthly, the above-mentioned objectives of this application are achieved through the following technical solutions:

[0059] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described UAV photovoltaic inspection and positioning method based on extended Kalman filtering.

[0060] In summary, this application includes at least one of the following beneficial technical effects:

[0061] 1. During the inspection of photovoltaic panels using drones, the system leverages multi-source sensors onboard the drone, including GPS, IMU, and vision sensors, to acquire real-time GPS positioning data, attitude data, and photovoltaic panel image data. By fusing absolute position information from GPS, high-frequency attitude data from the IMU, and image acquisition capabilities from the vision sensors, it effectively overcomes the signal obstruction problem caused by reliance on a single GPS source. This is particularly beneficial in mountainous or densely built-up areas, ensuring the comprehensiveness and reliability of the data source and laying a solid foundation for subsequent processing. In the stage of identifying faulty photovoltaic panels and extracting their pixel coordinates, the vision sensors perform real-time analysis of the image data, quickly detecting fault features such as hot spots, cracks, or obstructions. Edge detection algorithms accurately extract pixel coordinates, reducing the risk of subjective misjudgment inherent in traditional manual inspections and improving the accuracy and efficiency of fault identification. When converting the pixel coordinates of the faulty photovoltaic panels into initial position information in the geographic coordinate system, a transformation matrix is ​​established using real-time altitude, pitch, and yaw angle data from the drone, and calibrated using GPS data to generate a highly reliable initial position estimate. This solves the positioning deviation problem caused by inaccurate coordinate transformation in existing technologies. This provides high-quality input for subsequent filtering. In the stage of constructing the Extended Kalman Filter (EKF) model, the UAV's positioning data, attitude data, and initial position information obtained in the previous steps are used as input to establish a state vector containing the UAV's three-dimensional position, velocity, attitude angle, and coordinates of the faulty photovoltaic panel, as well as an observation vector composed of multiple source sensors. This design fully utilizes the advantages of the EKF in handling nonlinear systems. Through the state transition equation and observation equation, it achieves deep fusion of multi-source data, effectively compensating for data noise in complex environments (such as light interference or GPS signal loss), and significantly enhancing robustness and adaptability. In the process of iteratively calculating the optimal estimate and outputting the geographical coordinates of the faulty photovoltaic panel based on the state vector using the EKF algorithm, the EKF algorithm achieves centimeter-level accuracy in fault location output through state vector prediction in the prediction stage, linearization approximation of the Jacobian matrix, and Kalman gain calculation in the update stage. This ensures that fault information can be fed back to maintenance personnel in real time, which not only greatly improves the positioning accuracy to the centimeter level, but also enhances the stability in complex environments such as shading or mountainous areas. Ultimately, it reduces the operation and maintenance costs and fault response time of the photovoltaic power station, and improves power generation efficiency and system reliability.

[0062] 2. By preprocessing image data (eliminating illumination interference and background noise), image quality is significantly improved, solving visual interference problems in complex lighting environments (such as mountainous or shaded areas), reducing the risk of misjudgment due to shadows or reflections, and ensuring the reliability of subsequent analysis. By segmenting the preprocessed image data through temperature thresholding, abnormal temperature areas (such as hot spots or local overheating) are quickly extracted, fault points are directly located, and common photovoltaic panel faults (such as hot spot effect or aging) are efficiently identified. Based on the edge detection algorithm, the image contour boundary is extracted and a pixel coordinate system is generated, accurately delineating the geometric structure of the photovoltaic panel, avoiding the problems of blurred or discontinuous boundaries in traditional methods, providing a high-precision foundation for coordinate transformation, supporting centimeter-level positioning requirements, calculating the center pixel coordinates of the fault point and associating them with timestamp information, realizing synchronous recording of space and time, reducing data processing delay, and ensuring that fault location information is fed back to the positioning system in real time.

[0063] 3. Pixel coordinate distortion correction is performed (using the UAV's camera intrinsic parameter matrix and distortion coefficients to correct lens distortion), effectively eliminating barrel or pincushion distortion during image acquisition and solving coordinate offset problems caused by camera lens defects in complex environments. For example, it reduces pixel errors in mountainous or high-light scenes, providing high-fidelity input for subsequent conversions. When establishing the pixel coordinate to geographic coordinate conversion matrix, a mathematical model based on 3D spatial geometry is constructed by combining the UAV's real-time altitude, pitch angle, and yaw angle data. This accurately handles perspective distortion when the UAV is flying at an angle, ensuring consistent coordinate mapping under different viewpoints and avoiding scale distortion caused by attitude changes in traditional methods. The distortion-corrected pixel coordinates are mapped to the geographic coordinate system through a perspective transformation model, achieving a one-to-one correspondence between pixels and ground positions. Affine transformation or projection algorithms are used to efficiently process large-scale photovoltaic panel arrays. GPS positioning data is fused to calibrate the conversion results, compensating for accumulated errors caused by environmental interference (such as signal blockage), generating robust initial position information, and providing high-precision observation input for subsequent extended Kalman filtering.

[0064] 4. By integrating the initial position information of the faulty photovoltaic panel with the GPS positioning and attitude data of the UAV to generate the observation vector input, the system fully utilizes the complementarity of multi-source data (such as the absolute position accuracy of GPS, the high-frequency attitude update of IMU, and the local accuracy of visual initial positioning) to effectively overcome the data loss problem of a single sensor in obstructed or interfered environments, thus improving the system robustness. The UAV's three-dimensional position, velocity, attitude angle, and the geographical coordinates of the faulty photovoltaic panel are defined as a joint state vector, breaking through the limitation of traditional positioning methods that only focus on the UAV's own state. This achieves cooperative state estimation of the UAV and the faulty photovoltaic panel, significantly improving the fault location accuracy to the centimeter level. Based on the UAV's attitude data, a nonlinear state transition equation is constructed to accurately describe the UAV's kinematic laws and the evolution of the faulty photovoltaic panel's position over time. By using Taylor series linearization to approximate nonlinear systems, the system effectively suppresses state estimation drift in complex environments, ensuring the stability of photovoltaic panel fault location. The design of observation equations maps the joint state vector to multi-source sensor data and initial position information, forming a closed-loop feedback mechanism. Kalman gain is used to dynamically adjust the state estimation weights, correcting positioning errors in real time. Simultaneously, efficient data compression and parallel computing optimize the processing flow, reducing positioning latency and providing maintenance personnel with real-time and accurate fault location information. The extended Kalman filter model, through joint definition of state vectors (covering both UAVs and photovoltaic panels) and multi-source fusion of observation equations (integrating GPS / IMU / visual data), maintains centimeter-level positioning accuracy even in complex scenarios such as mountainous areas with obstructions, significantly improving the efficiency of photovoltaic power plant operation and maintenance and reducing maintenance costs. Attached Figure Description

[0065] Figure 1 This is a flowchart of a UAV photovoltaic inspection and positioning method based on extended Kalman filtering in one embodiment of this application;

[0066] Figure 2 This is a flowchart illustrating the implementation of step S20 in a UAV photovoltaic inspection and positioning method based on extended Kalman filtering in one embodiment of this application.

[0067] Figure 3 This is a flowchart illustrating the implementation of step S30 in a UAV photovoltaic inspection and positioning method based on extended Kalman filtering in one embodiment of this application.

[0068] Figure 4 This is a flowchart illustrating the implementation of step S40 in a UAV photovoltaic inspection and positioning method based on extended Kalman filtering in one embodiment of this application.

[0069] Figure 5 This is a flowchart illustrating the implementation of step S50 in a UAV photovoltaic inspection and positioning method based on extended Kalman filtering in one embodiment of this application.

[0070] Figure 6 This is a flowchart illustrating the implementation of step S52 in a UAV photovoltaic inspection and positioning method based on extended Kalman filtering in one embodiment of this application.

[0071] Figure 7 This is another implementation flowchart of a UAV photovoltaic inspection and positioning method based on extended Kalman filtering in one embodiment of this application;

[0072] Figure 8 This is a principle block diagram of a UAV photovoltaic inspection and positioning system based on extended Kalman filtering in one embodiment of this application;

[0073] Figure 9 This is a schematic diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0074] The present application will be further described in detail below with reference to the accompanying drawings.

[0075] In one embodiment, such as Figure 1 As shown, this application discloses a UAV photovoltaic inspection and positioning method based on extended Kalman filtering, which specifically includes the following steps:

[0076] S10: Real-time acquisition of GPS positioning data, attitude data, and photovoltaic panel image data of the drone based on the sensors carried by the drone.

[0077] Specifically, the GPS positioning data, attitude data, and photovoltaic panel image data of the drone are acquired in real time and synchronously through the GPS module, IMU (inertial measurement unit), and high-definition vision sensor (including infrared thermal imager) carried by the drone.

[0078] Furthermore, by integrating absolute position information from GPS, high-frequency attitude data from IMU, and image acquisition capabilities from visual sensors, the signal obstruction problem caused by reliance on a single GPS source is effectively overcome. This is particularly useful in mountainous or densely built-up areas, ensuring the comprehensiveness and reliability of the data source and laying a solid foundation for subsequent processing.

[0079] S20: Identify the fault characteristics of the photovoltaic panel based on the image data of the photovoltaic panel and extract its pixel coordinates.

[0080] Specifically, in the stage of identifying the fault characteristics of the photovoltaic panel and extracting its pixel coordinates, a visual sensor is used to analyze the image data in real time to quickly detect fault characteristics such as hot spot effect, cracks or shading, and the pixel coordinates are accurately extracted through edge detection algorithm, which reduces the risk of subjective misjudgment in traditional manual inspection and improves the accuracy and efficiency of fault identification.

[0081] S30: Convert the pixel coordinates of the faulty photovoltaic panel into initial location information in the geographic coordinate system.

[0082] Specifically, when converting the pixel coordinates of the faulty photovoltaic panel into initial position information in the geographic coordinate system, a transformation matrix is ​​established by combining the real-time altitude, pitch angle, and yaw angle data of the UAV, and calibrated by GPS data to generate a highly reliable initial position estimate. This solves the positioning deviation problem caused by inaccurate coordinate transformation in the existing technology and provides high-quality input for subsequent filtering processing.

[0083] S40: Construct an extended Kalman filter model by inputting the UAV's positioning data, attitude data, and initial position information into the extended Kalman filter model to establish state vectors and observation vectors.

[0084] Specifically, in the stage of constructing the extended Kalman filter model, the UAV's positioning data, attitude data, and initial position information obtained in the aforementioned steps are used as inputs to establish a state vector containing the UAV's three-dimensional position, velocity, attitude angle, and coordinates of the faulty photovoltaic panel, as well as an observation vector composed of multiple source sensors. This design fully utilizes the advantages of extended Kalman filtering in handling nonlinear systems, and achieves deep fusion of multi-source data through state transition equations and observation equations, effectively compensating for data noise in complex environments (such as light interference or GPS signal loss), and significantly enhancing robustness and adaptability.

[0085] S50: Based on the state vector, the optimal estimate is calculated iteratively using the extended Kalman filter algorithm, and the geographical coordinates of the faulty photovoltaic panel are output.

[0086] Specifically, in the process of iteratively calculating the optimal estimate and outputting the geographical coordinates of the faulty photovoltaic panel based on the state vector using the extended Kalman filter algorithm, the EKF algorithm achieves centimeter-level accuracy in outputting the fault location by predicting the state vector in the prediction stage, linearizing the Jacobian matrix, and calculating the Kalman gain in the update stage, ensuring that fault information can be fed back to maintenance personnel in real time.

[0087] In this embodiment, during the inspection of photovoltaic panels using a drone, the drone's multi-source sensors, including GPS, IMU, and vision sensors, are used to acquire real-time GPS positioning data, attitude data, and photovoltaic panel image data. By fusing the absolute position information of GPS, the high-frequency attitude data of IMU, and the image acquisition capabilities of the vision sensor, the signal obstruction problem caused by reliance on a single GPS source is effectively overcome. This is particularly beneficial in mountainous or densely built-up areas, ensuring the comprehensiveness and reliability of the data source and laying a solid foundation for subsequent processing. In the stage of identifying the fault characteristics of the photovoltaic panel and extracting its pixel coordinates, the vision sensor is used to analyze the image data in real time, quickly detecting fault characteristics such as hot spot effects, cracks, or obstructions. An edge detection algorithm is used to accurately extract pixel coordinates, reducing the risk of subjective misjudgment in traditional manual inspections and improving the accuracy and efficiency of fault identification. When converting the pixel coordinates of the faulty photovoltaic panel into initial position information in the geographic coordinate system, a transformation matrix is ​​established by combining the drone's real-time altitude, pitch angle, and yaw angle data, and calibrated using GPS data to generate a highly reliable initial position estimate. This solves the positioning deviation problem caused by inaccurate coordinate transformation in existing technologies. This provides high-quality input for subsequent filtering. In the stage of constructing the Extended Kalman Filter (EKF) model, the UAV's positioning data, attitude data, and initial position information obtained in the previous steps are used as input to establish a state vector containing the UAV's three-dimensional position, velocity, attitude angle, and coordinates of the faulty photovoltaic panel, as well as an observation vector composed of multiple source sensors. This design fully utilizes the advantages of the EKF in handling nonlinear systems, achieving deep fusion of multi-source data through state transition equations and observation equations. This effectively compensates for data noise in complex environments (such as light interference or GPS signal loss), significantly enhancing robustness and adaptability. In the process of iteratively calculating the optimal estimate and outputting the geographical coordinates of the faulty photovoltaic panel based on the state vector using the EKF algorithm, the EKF algorithm achieves centimeter-level accuracy in fault location output through state vector prediction in the prediction stage, linearization approximation of the Jacobian matrix, and Kalman gain calculation in the update stage. This ensures that fault information can be fed back to maintenance personnel in real time, not only significantly improving positioning accuracy to the centimeter level but also enhancing stability in complex environments such as shaded or mountainous areas. Ultimately, this reduces the operation and maintenance costs and fault response time of the photovoltaic power station, and improves power generation efficiency and system reliability.

[0088] In one embodiment, such as Figure 2 As shown, in step S20, which involves identifying the fault characteristics of the photovoltaic panel based on its image data and extracting its pixel coordinates, the specific steps include:

[0089] S21: Preprocess the image data of the photovoltaic panel to eliminate light interference and background noise, and obtain preprocessed photovoltaic panel image data.

[0090] Specifically, by preprocessing image data (eliminating illumination interference and background noise), image quality is significantly improved, visual interference problems in complex lighting environments (such as mountainous or occluded areas) are solved, the risk of misjudgment caused by shadows or reflections is reduced, and the reliability of subsequent analysis is ensured.

[0091] S22: Perform temperature threshold segmentation on the preprocessed photovoltaic panel image data, extract abnormal temperature areas, and locate fault points.

[0092] Specifically, by segmenting the preprocessed image data using temperature thresholds, abnormal temperature areas (such as hot spots or localized overheating) can be quickly extracted, fault points can be directly located, and common photovoltaic panel faults (such as hot spot effects or aging) can be efficiently identified.

[0093] S23: Extract the contour boundary of the preprocessed photovoltaic panel image data based on the edge detection algorithm, and generate the pixel coordinate system of the photovoltaic panel image data according to the contour boundary.

[0094] Specifically, the edge detection algorithm extracts the image contour boundary and generates a pixel coordinate system to accurately delineate the geometric structure of the photovoltaic panel, avoiding the problems of blurred or discontinuous boundaries in traditional methods, providing a high-precision foundation for coordinate transformation, and supporting centimeter-level positioning requirements.

[0095] S24: Calculate the pixel coordinates of the fault point center in the preprocessed photovoltaic panel image data and associate it with timestamp information.

[0096] Specifically, the pixel coordinates of the fault point's center are calculated and associated with timestamp information to achieve synchronized spatial and temporal recording, reducing data processing latency and ensuring that fault location information is fed back to the positioning system in real time.

[0097] Furthermore, these steps work together to not only significantly improve the accuracy of fault identification, but also enhance the system's stability in complex environments by optimizing the image processing flow (such as reducing latency through parallel computing), providing high-quality input for subsequent extended Kalman filtering, and ultimately improving the real-time performance, reliability, and intelligence of photovoltaic power plant fault location, thereby reducing operation and maintenance costs.

[0098] In one embodiment, such as Figure 3 As shown, in step S30, the pixel coordinates of the faulty photovoltaic panel are converted into initial location information in the geographic coordinate system, specifically including:

[0099] S31: Perform distortion correction on pixel coordinates based on the intrinsic parameter matrix and distortion coefficients of the UAV's camera.

[0100] Specifically, by performing distortion correction on pixel coordinates, such as using the camera intrinsic matrix and distortion coefficients of the drone to correct lens distortion, barrel or pincushion distortion during image acquisition can be effectively eliminated, and coordinate offset problems caused by camera lens defects in complex environments can be solved. For example, in mountainous or strong light scenes, pixel errors can be reduced by up to 40%, providing high-fidelity input for subsequent conversion.

[0101] S32: Combine the real-time altitude, pitch angle and yaw angle data of the UAV to establish a transformation matrix from pixel coordinates to geographic coordinates.

[0102] Specifically, when establishing the transformation matrix from pixel coordinates to geographic coordinates, a mathematical model based on three-dimensional spatial geometry is constructed by combining the real-time altitude, pitch angle, and yaw angle data of the UAV (such as attitude information provided by IMU sensors). This model accurately handles perspective distortion when the UAV is flying at an angle, ensuring the consistency of coordinate mapping under different viewpoints, avoiding scale distortion caused by attitude changes in traditional methods, and improving transformation efficiency.

[0103] S33: Map the distorted pixel coordinates to the geographic coordinate system using a perspective transformation model.

[0104] Specifically, the perspective transformation model maps the distortion-free pixel coordinates to the geographic coordinate system, achieving a one-to-one correspondence between pixel points and ground locations. Affine transformation or projection algorithms are used to efficiently process large-scale photovoltaic arrays, reducing computational latency and supporting centimeter-level positioning requirements.

[0105] S34: Merge GPS positioning data to perform position calibration on the conversion results and generate initial position information.

[0106] Specifically, the conversion results are calibrated by integrating GPS positioning data, such as through weighted averaging or bias compensation, to compensate for cumulative errors caused by environmental interference (such as signal blockage), generating robust initial position information and providing high-precision observation input for subsequent extended Kalman filtering.

[0107] In one embodiment, such as Figure 4 As shown, in step S40, the extended Kalman filter model is constructed by inputting the UAV's positioning data, attitude data, and initial position information into the extended Kalman filter model to establish the state vector and observation vector. Specifically, this includes:

[0108] S41: Generate observation vector input based on the initial position information of the faulty photovoltaic panel and the GPS positioning data and attitude data of the UAV.

[0109] Specifically, by integrating the initial position information of the faulty photovoltaic panel with the GPS positioning data and attitude data of the UAV to generate the observation vector input, the system makes full use of the complementarity of multi-source data (such as the absolute position accuracy of GPS, the high-frequency attitude update of IMU, and the local accuracy of visual initial positioning) to effectively overcome the data loss problem of a single sensor in the environment of obstruction or interference, and improve the robustness of the system.

[0110] S42: Define the UAV's three-dimensional position, velocity, attitude angle, and the geographic coordinates of the faulty photovoltaic panel to be located as a joint state vector.

[0111] Specifically, the three-dimensional position, velocity, attitude angle of the UAV and the geographical coordinates of the faulty photovoltaic panel are defined as a joint state vector. This breaks through the limitation of traditional positioning methods that only focus on the state of the UAV itself, and realizes the collaborative state estimation of the UAV and the faulty photovoltaic panel, significantly improving the fault location accuracy to the centimeter level.

[0112] S43: Construct a nonlinear state transition equation based on the attitude data of the UAV to describe the evolution of the joint state vector over time.

[0113] Specifically, a nonlinear state transition equation is constructed based on the attitude data of the UAV to accurately describe the kinematic laws of the UAV and the evolution of the position of the faulty photovoltaic panel over time. The nonlinear system is approximated by Taylor series linearization, which effectively suppresses the state estimation drift in complex environments (such as strong wind disturbances or maneuvering flight) and ensures the stability of the model.

[0114] S44: Map the joint state vector to the initial position information of the multi-source sensors and the faulty photovoltaic panel carried by the UAV to form the observation equation.

[0115] Specifically, the design of the observation equation maps the joint state vector to multi-source sensor data and initial position information, forming a closed-loop feedback mechanism. The state estimation weights are dynamically adjusted through Kalman gain to correct positioning deviations in real time. At the same time, the efficient data compression and parallel computing are combined to optimize the processing flow, reducing positioning latency and providing maintenance personnel with real-time and accurate fault location information.

[0116] Furthermore, the extended Kalman filter model maintains centimeter-level positioning accuracy even in complex scenarios such as mountainous areas with obstructions by jointly defining state vectors (covering dual targets of UAVs and photovoltaic panels) and multi-source fusion of observation equations (integrating GPS / IMU / visual data), thus improving positioning stability compared to existing technologies.

[0117] In one embodiment, such as Figure 5 As shown, in step S50, the optimal estimate is iteratively calculated using the extended Kalman filter algorithm based on the state vector, and the geographical coordinates of the faulty photovoltaic panel are output. Specifically, this includes:

[0118] S51: Predict the state vector and covariance matrix at the next time step based on the nonlinear state transition equation.

[0119] S52: Linearize the nonlinear state transition equation and solve for the state transition matrix and observation matrix.

[0120] S53: Calculate the Kalman gain by combining sensor observations, and update the state vector and covariance matrix.

[0121] S54: Iterate through the above steps until the state estimation error converges to a centimeter-level threshold.

[0122] Specifically, based on the nonlinear state transition equation, the system predicts the state vector and covariance matrix of the next moment. It utilizes the UAV kinematic model (such as the correlation between the rate of change of attitude angle and velocity) to predict the state evolution trend, effectively compensating for positioning interruptions caused by GPS signal loss or obstruction. It maintains predictive stability even in complex environments such as mountainous areas with strong winds. By linearizing the nonlinear state transition equation (using Taylor series expansion to solve the Jacobian matrix), nonlinear variables such as the UAV attitude angle and the position of the faulty photovoltaic panel are transformed into a computable linear system. This overcomes the limitations of traditional linear filtering algorithms in scenarios with tilted photovoltaic panel installations, improving positioning accuracy to the centimeter level. Then, it combines sensor observations (GPS / IMU / initial position information)... The system calculates the Kalman gain in real time and dynamically adjusts the state estimation weights (such as increasing the weight of the initial visual position data when the GPS signal is weak). It achieves closed-loop feedback calibration of multi-source data by updating the state vector and covariance matrix, which significantly suppresses data noise interference (such as IMU drift or image coordinate transformation deviation) and compresses the fluctuation range of the positioning result to within the centimeter level. The above steps are iteratively executed until the state estimation error converges to the centimeter-level threshold, ensuring that the output result meets the requirements of photovoltaic panel-level accurate positioning. The overall technical solution uses a closed-loop iterative mechanism of prediction-linearization-update-convergence to significantly enhance the system's adaptability in occlusion, strong light, and dynamic flight scenarios while ensuring centimeter-level positioning accuracy, and reduces fault location delay.

[0123] In one embodiment, such as Figure 6 As shown, in step S52, the nonlinear state transition equation is linearized and approximated, specifically including:

[0124] S521: Solve for the Jacobian matrix using the first-order partial derivatives of the state transition equation through Taylor series expansion.

[0125] S522: Correlate the rate of change of the coordinates of the faulty photovoltaic panel with the rate of change of the attitude angle of the UAV.

[0126] S523: Compensates for state prediction errors when GPS signals are lost by using attitude data.

[0127] S524: Quantifying sensor data uncertainty based on process noise covariance matrix.

[0128] Specifically, by using Taylor series expansion to solve the first-order partial derivatives of the state transition equation to generate the Jacobian matrix, nonlinear variables such as UAV attitude angles and the position of faulty photovoltaic panels are transformed into computable linear systems (e.g., relating the pitch angle change rate to the photovoltaic panel coordinate derivative). This overcomes the limitations of traditional algorithms in scenarios with tilted photovoltaic panel installations, improves model computational efficiency, and dynamically correlates the faulty photovoltaic panel coordinate change rate with the UAV attitude angle change rate (e.g., correcting coordinate offsets using the roll rate obtained in real-time from the IMU). This accurately describes the impact of complex maneuvering flight (e.g., sharp turns during mountain inspections) on positioning, solves the position drift problem in dynamic scenarios, and mitigates attitude interference. The resulting error is reduced by compensating for state prediction errors when GPS signals are lost through attitude data (e.g., using IMU angular velocity integral to replace GPS positioning in obscured areas), maintaining system stability through the prediction mechanism of extended Kalman filtering, ensuring that positioning accuracy remains at the centimeter level even when GPS is interrupted for a certain period of time, quantifying sensor data uncertainty based on the process noise covariance matrix (e.g., assigning differentiated noise weights to GPS / IMU / visual data), and adaptively adjusting the multi-source data fusion ratio through Kalman gain to suppress sensor noise interference (e.g., IMU zero bias or residual image distortion error), thus compressing the fluctuation range of positioning results to within the centimeter level.

[0129] In one embodiment, such as Figure 7 As shown, after step S50, that is, after iteratively calculating the optimal estimate based on the state vector using the extended Kalman filter algorithm and outputting the geographical coordinates of the faulty photovoltaic panel, the UAV photovoltaic inspection and positioning method based on extended Kalman filter further includes:

[0130] S60: Map the geographic coordinates of the extended Kalman filter output to the photovoltaic panel array number to generate an inspection report containing the fault location, fault type and confidence level;

[0131] S70: Based on the wireless network, the report is pushed to the operation and maintenance terminal in real time, and the coordinates of the fault point are dynamically marked in the drone flight path planning.

[0132] Specifically, by accurately mapping geographic coordinates to photovoltaic panel array numbers (such as rasterized matching based on the power plant GIS database), the problem of the disconnect between traditional positioning results and physical location is solved. This allows maintenance personnel to directly locate the faulty panel number and generate a multi-dimensional inspection report containing the fault location, fault type (such as hot spot / crack / shading), and confidence level (based on EKF covariance matrix calculation). By integrating infrared thermal imaging data with visual analysis results, intelligent diagnosis of fault nature is achieved. The report is pushed to the maintenance terminal (such as mobile device or central control system) in real time via 4G / 5G wireless network. The fault point coordinates are dynamically marked in the drone flight path planning (such as automatic obstacle avoidance or close-range re-inspection). The optimal inspection path is constructed by combining the real-time updated digital map.

[0133] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0134] In one embodiment, a UAV photovoltaic inspection and positioning system based on extended Kalman filtering is provided. This UAV photovoltaic inspection and positioning system based on extended Kalman filtering corresponds one-to-one with the UAV photovoltaic inspection and positioning method based on extended Kalman filtering in the above embodiments. Figure 8 As shown, the UAV photovoltaic inspection and positioning system based on extended Kalman filtering includes a data acquisition module, a photovoltaic panel fault identification module, a coordinate transformation module, a model building module, and a photovoltaic panel fault coordinate output module. Detailed descriptions of each functional module are as follows:

[0135] The data acquisition module is used to acquire GPS positioning data, attitude data, and photovoltaic panel image data of the UAV in real time based on the sensors carried by the UAV.

[0136] A photovoltaic panel fault identification module is used to identify the fault characteristics of a faulty photovoltaic panel based on the image data of the photovoltaic panel and extract its pixel coordinates;

[0137] The coordinate transformation module is used to convert the pixel coordinates of the faulty photovoltaic panel into initial location information in the geographic coordinate system;

[0138] The model building module is used to build an extended Kalman filter model. The positioning data, attitude data and initial position information of the UAV are input into the extended Kalman filter model to establish the state vector and observation vector.

[0139] The photovoltaic panel fault coordinate output module is used to iteratively calculate the optimal estimate based on the state vector using the extended Kalman filter algorithm, and output the geographical coordinates of the faulty photovoltaic panel.

[0140] Specific limitations regarding the UAV photovoltaic inspection and positioning system based on extended Kalman filtering can be found in the limitations of the UAV photovoltaic inspection and positioning method based on extended Kalman filtering mentioned above, and will not be repeated here. Each module in the aforementioned UAV photovoltaic inspection and positioning system based on extended Kalman filtering can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of the electronic device in hardware form or independently of the processor, or stored in the memory of the electronic device in software form, so that the processor can call and execute the corresponding operations of each module.

[0141] In one embodiment, an electronic device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 9 As shown, this electronic device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage medium. The database stores the database. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a UAV photovoltaic inspection and positioning method based on extended Kalman filtering.

[0142] In one embodiment, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps:

[0143] The drone's GPS positioning data, attitude data, and photovoltaic panel image data are acquired in real time using sensors mounted on the drone.

[0144] Based on the image data of the photovoltaic panel, identify the fault characteristics of the faulty photovoltaic panel and extract its pixel coordinates;

[0145] Convert the pixel coordinates of the faulty photovoltaic panel into initial location information in the geographic coordinate system;

[0146] An extended Kalman filter model is constructed by inputting the UAV's positioning data, attitude data, and initial position information into the extended Kalman filter model to establish state vectors and observation vectors.

[0147] The optimal estimate is calculated iteratively using the extended Kalman filter algorithm based on the state vector, and the geographical coordinates of the faulty photovoltaic panel are output.

[0148] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0149] The drone's GPS positioning data, attitude data, and photovoltaic panel image data are acquired in real time using sensors mounted on the drone.

[0150] Based on the image data of the photovoltaic panel, identify the fault characteristics of the faulty photovoltaic panel and extract its pixel coordinates;

[0151] Convert the pixel coordinates of the faulty photovoltaic panel into initial location information in the geographic coordinate system;

[0152] An extended Kalman filter model is constructed by inputting the UAV's positioning data, attitude data, and initial position information into the extended Kalman filter model to establish state vectors and observation vectors.

[0153] The optimal estimate is calculated iteratively using the extended Kalman filter algorithm based on the state vector, and the geographical coordinates of the faulty photovoltaic panel are output.

[0154] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0155] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0156] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for unmanned aerial vehicle (UAV) photovoltaic inspection and positioning based on extended Kalman filtering, characterized in that, The UAV photovoltaic inspection and positioning method based on extended Kalman filtering includes the following steps: The drone's GPS positioning data, attitude data, and photovoltaic panel image data are acquired in real time using sensors mounted on the drone. Based on the image data of the photovoltaic panel, identify the fault characteristics of the faulty photovoltaic panel and extract its pixel coordinates; Convert the pixel coordinates of the faulty photovoltaic panel into initial location information in the geographic coordinate system; An extended Kalman filter model is constructed by inputting the UAV's positioning data, attitude data, and initial position information into the extended Kalman filter model to establish state vectors and observation vectors. The optimal estimate is calculated iteratively using the extended Kalman filter algorithm based on the state vector, and the geographical coordinates of the faulty photovoltaic panel are output.

2. The UAV photovoltaic inspection and positioning method based on extended Kalman filtering according to claim 1, characterized in that, The step of identifying the fault characteristics of a faulty photovoltaic panel and extracting its pixel coordinates based on the image data of the photovoltaic panel specifically includes: The image data of the photovoltaic panel is preprocessed to eliminate light interference and background noise, resulting in preprocessed photovoltaic panel image data. Temperature threshold segmentation is performed on the preprocessed photovoltaic panel image data to extract abnormal temperature areas and locate fault points. The contour boundaries of the preprocessed photovoltaic panel image data are extracted based on the edge detection algorithm, and the pixel coordinate system of the photovoltaic panel image data is generated based on the contour boundaries. Calculate the pixel coordinates of the fault point center in the preprocessed photovoltaic panel image data and associate them with timestamp information.

3. The UAV photovoltaic inspection and positioning method based on extended Kalman filtering according to claim 1, characterized in that, The process of converting the pixel coordinates of the faulty photovoltaic panel into initial location information in a geographic coordinate system specifically includes: Based on the intrinsic parameter matrix and distortion coefficients of the drone's camera, the pixel coordinates are subjected to distortion correction. By combining the real-time altitude, pitch angle, and yaw angle data of the UAV, a transformation matrix from pixel coordinates to geographic coordinates is established; The distortion-free pixel coordinates are mapped to the geographic coordinate system using a perspective transformation model; The conversion results are calibrated by integrating GPS positioning data to generate initial location information.

4. The UAV photovoltaic inspection and positioning method based on extended Kalman filtering according to claim 1, characterized in that, The construction of the extended Kalman filter model involves inputting the UAV's positioning data, attitude data, and initial position information into the extended Kalman filter model to establish state vectors and observation vectors. Specifically, this includes: An observation vector input is generated based on the initial position information of the faulty photovoltaic panel and the GPS positioning and attitude data of the UAV; The three-dimensional position, velocity, attitude angle of the UAV, and the geographic coordinates of the faulty photovoltaic panel to be located are defined as a joint state vector; A nonlinear state transition equation is constructed based on the attitude data of the UAV to describe the evolution of the joint state vector over time. The joint state vector is mapped to the initial position information of the multi-source sensors and the faulty photovoltaic panel carried by the UAV to form the observation equation.

5. The UAV photovoltaic inspection and positioning method based on extended Kalman filtering according to claim 4, characterized in that, The step of iteratively calculating the optimal estimate based on the state vector using the extended Kalman filter algorithm and outputting the geographical coordinates of the faulty photovoltaic panel specifically includes: Predict the state vector and covariance matrix at the next time step based on the nonlinear state transition equation; The nonlinear state transition equation is linearized and approximated to solve for the state transition matrix and the observation matrix. Calculate the Kalman gain by combining sensor observations, and update the state vector and covariance matrix; Repeat the above steps iteratively until the state estimation error converges to a centimeter-level threshold.

6. The UAV photovoltaic inspection and positioning method based on extended Kalman filtering according to claim 5, characterized in that, The linearization approximation of the nonlinear state transition equation specifically includes: The Jacobian matrix is ​​solved by using the first-order partial derivatives of the state transition equation through Taylor series expansion; Correlate the rate of change of the coordinates of the faulty photovoltaic panel with the rate of change of the drone's attitude angle; Compensate for state prediction errors when GPS signals are lost by using attitude data; Quantification of sensor data uncertainty based on process noise covariance matrix.

7. The UAV photovoltaic inspection and positioning method based on extended Kalman filtering according to claim 1, characterized in that, After iteratively calculating the optimal estimate based on the state vector using the extended Kalman filter algorithm and outputting the geographical coordinates of the faulty photovoltaic panel, the UAV photovoltaic inspection and positioning method based on extended Kalman filter further includes: The geographic coordinates output by the extended Kalman filter are mapped to the photovoltaic panel array number to generate an inspection report containing the fault location, fault type, and confidence level. The report is pushed to the operation and maintenance terminal in real time via wireless network, and the coordinates of the fault point are dynamically marked in the drone flight path planning.

8. A UAV photovoltaic inspection and positioning system based on extended Kalman filtering, characterized in that, The UAV photovoltaic inspection and positioning system based on extended Kalman filtering includes: The data acquisition module is used to acquire GPS positioning data, attitude data, and photovoltaic panel image data of the UAV in real time based on the sensors carried by the UAV. A photovoltaic panel fault identification module is used to identify the fault characteristics of a faulty photovoltaic panel based on the image data of the photovoltaic panel and extract its pixel coordinates; The coordinate transformation module is used to convert the pixel coordinates of the faulty photovoltaic panel into initial location information in the geographic coordinate system; The model building module is used to build an extended Kalman filter model. The positioning data, attitude data and initial position information of the UAV are input into the extended Kalman filter model to establish the state vector and observation vector. The photovoltaic panel fault coordinate output module is used to iteratively calculate the optimal estimate based on the state vector using the extended Kalman filter algorithm, and output the geographical coordinates of the faulty photovoltaic panel.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the UAV photovoltaic inspection and positioning method based on extended Kalman filtering as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the UAV photovoltaic inspection and positioning method based on extended Kalman filtering as described in any one of claims 1 to 7.

Images