Method and system for constructing point cloud map of electric power operation based on inspection robot

Abstract

The present invention relates to the technical field of video measurement, and in particular to a method and system for constructing a point cloud map of electric power operations based on an inspection robot. The method comprises: acquiring multi-source sensor data of an electric power operation environment; synchronously calibrating the acquired multi-source sensor data, and verifying the time consistency of the multi-source sensor data; on the premise that the time consistency verification is passed, performing point cloud registration using an improved point cloud registration algorithm to obtain an optimized weight matrix; performing posture estimation on the inspection robot based on the posture data from an inertial measurement unit, and calculating the posture data of the inspection robot; fusing the optimized weight matrix with the posture data of the inspection robot, and filtering and updating the fused point cloud matrix using a filtering algorithm; and constructing an electric power operation point cloud map of an electric power operation environment based on the filtered and updated point cloud matrix. The present invention can provide accurate three-dimensional visualization support for electric power operations, and improve the efficiency and safety of operations.

Description

Technical Field

[0001] The present invention relates to the technical field of video metrology, and particularly to a method and system for constructing a power operation point cloud map based on an inspection robot. Background Art

[0002] With the complexity of the power system and the continuous expansion of its scale, the stable operation of power equipment has become particularly important. Traditional power inspections mainly rely on manual inspections, which are not only time-consuming and laborious, but also have problems such as high human resource costs, low inspection efficiency, and high risks. To improve inspection efficiency and safety, inspection robots have emerged. Through autonomous navigation and intelligent perception technologies, inspection robots can accurately locate power equipment, monitor its status, and detect faults, greatly reducing the workload and risks of manual inspections.

[0003] In power operations, constructing an accurate operation point cloud map is the basis for the efficient operation of inspection robots. The point cloud map can provide accurate environmental information for the robot, helping it to complete navigation, obstacle detection, and operation planning. Based on the point cloud map, inspection robots can accurately locate the position of power equipment and achieve efficient automatic inspection operations.

[0004] The power operation environment is complex and changeable, including a large number of power equipment, lines, and complex terrains. Traditional point cloud construction methods are difficult to meet the requirements of these specific environments. The precise positioning and identification of power equipment require the point cloud map to have a high resolution and accuracy, which poses higher requirements for existing point cloud processing algorithms. The real-time and safety requirements of power inspection tasks require the construction and update speed of the point cloud map to be fast enough to support the real-time operation of inspection robots.

[0005] Although certain progress has been made in point cloud map construction in the prior art, there are still many technical problems to be solved in the power operation scenario, such as the acquisition and processing efficiency of the point cloud matrix, environmental adaptability, and equipment positioning accuracy. Therefore, in view of the specific requirements of power operations, developing a method and system for constructing a power operation point cloud map based on an inspection robot has important practical significance and application value.

[0006] The technical routes of the prior art mainly include:

[0007] 1. SLAM technology based on lidar: The environment is scanned by a lidar (LiDAR) sensor to obtain a three-dimensional point cloud matrix, and the SLAM (Simultaneous Localization and Mapping) algorithm is used to perform robot positioning and environmental map construction simultaneously. Lidar SLAM is widely used in unmanned driving and robot navigation.

[0008] Existing problems: For complex power operation environments, the resolution and scanning accuracy of lidar may be insufficient. Especially in the environment of power equipment erected at high altitudes and dense power lines, data loss or accuracy degradation is likely to occur. The SLAM algorithm has high real-time requirements in power operations, but in complex environments, its computational efficiency may be insufficient, affecting the efficiency of real-time navigation and operations.

[0009] 2. 3D reconstruction technology based on depth cameras: Depth cameras are used to capture the depth information of the environment, and point cloud maps are generated through 3D reconstruction algorithms for robot navigation and power equipment identification. Depth cameras are usually combined with RGB cameras to provide richer environmental information.

[0010] Existing problems: The performance of depth cameras is unstable under strong light, low light, or adverse weather conditions, which may lead to a decline in the quality of the point cloud matrix, affecting the accuracy of map construction. When 3D reconstruction algorithms process complex power equipment, they are easily affected by occlusion and reflection, resulting in errors in the reconstruction results and affecting the precise positioning and identification of equipment.

[0011] 3. Point cloud map construction based on multi-sensor fusion: Multiple sensors such as lidar, depth cameras, and inertial measurement units (IMUs) are comprehensively used, and data fusion technology is used to improve the accuracy and robustness of the point cloud map. Sensor fusion technology can complement each other in different environments, enhancing the map construction effect.

[0012] Existing problems: The data synchronization and fusion algorithms of multi-sensors are complex, and problems such as timing errors and data inconsistency are likely to occur, affecting the accuracy of the point cloud map. The diversity of sensors increases the hardware cost and computational complexity of the system. Especially in scenarios where power inspection robots need to operate stably for a long time, the maintenance and debugging are more difficult. Summary of the Invention

[0013] To achieve the objectives of the present invention, the present invention is implemented through the following technical solutions:

[0014] A method for constructing a point cloud map for power operations based on an inspection robot, the point cloud map is formed in the form of a point cloud matrix, and the method includes the following steps:

[0015] S1. Obtain multi-source sensor data of the power operation environment, the multi-source sensor data includes a basic point cloud matrix from a lidar, a registered point cloud matrix from a depth camera, and pose data from an inertial measurement unit, and the pose data includes position data and attitude data;

[0016] S2. Synchronously calibrate the obtained multi-source sensor data to verify the time consistency of the multi-source sensor data;

[0017] S3. On the premise that the time consistency check passes, use an improved point cloud registration algorithm to perform point cloud registration on the basic point cloud matrix from the lidar and the registered point cloud matrix from the depth camera to obtain an optimized weight matrix;

[0018] S4. Use the Kalman filtering algorithm to perform pose estimation on the inspection robot based on the pose data from the inertial measurement unit, and calculate the pose data of the inspection robot;

[0019] S5. Integrate the optimized weight matrix and the pose data of the inspection robot in steps S3 and S4, and use a filtering algorithm to filter and update the integrated point cloud matrix;

[0020] S6. Construct a power operation point cloud map of the power operation environment based on the filtered and updated point cloud matrix.

[0021] Further, the improved point cloud registration algorithm includes:

[0022] S31. Select the nearest points in the basic point cloud matrix and the registered point cloud matrix, and calculate the weight of each point. The weight is calculated by the following formula:

[0023] ,

[0024] where, represents the weight of the i th point; is the dynamic adjustment parameter of the i th point, and the initial value is 1; represents the i th point in the basic point cloud matrix, represents the i th point in the registered point cloud matrix with the same serial number as that in the basic point cloud matrix, represents the natural exponential function, and the coordinate parameters of point and point satisfy:

[0025] ,

[0026] where, both the basic point cloud matrix and the registered point cloud matrix are expressed based on the OXYZ orthogonal coordinate system, represents the X-axis coordinate of point , represents the Y-axis coordinate of point , represents the Z-axis coordinate of point , represents the X-axis coordinate of point , represents the X-axis of point ​ Represents the Z-axis coordinate of a point ,

[0027] Represents the Euclidean distance between a point and a point , satisfying:

[0028] ,

[0029] S32. Iteratively optimize the initial weight matrix using the weighted least squares method according to the weight of each point to obtain an optimized weight matrix, satisfying:

[0030] ,

[0031] wherein, represents the optimized weight matrix; represents the initial weight matrix; the initial weight matrix and the optimized weight matrix each element of which represents the weight value of the corresponding point; the initial values of all elements of the initial weight matrix are 1 and are updated after each iterative optimization;

[0032] represents the maximum value of the number of points in the optimized weight matrix ; represents the iterative optimization function; represents the minimization value; represents the optimal value of the weight that makes the iterative optimization function obtain the minimum value through iterative calculation, and is used as the optimal weight value , and the optimized weight matrix is composed of all the optimal weight values .

[0033] Furthermore, in step S31, the dynamic adjustment parameter i of the th point is adjusted according to the density and dynamic adjustment of the base point cloud matrix and the registered point cloud matrix, including: for the i th point, adjust the dynamic adjustment parameter i of the i th point according to the distribution density of the neighborhood points of the th point, satisfying:

[0034] ,

[0035] wherein, represents the i th point in the base point cloud matrix, Indicates the i th point in the registered point cloud matrix with the same serial number as that in the base point cloud matrix, represents the j th point in the base point cloud matrix, indicates the j th point in the registered point cloud matrix with the same serial number as that in the base point cloud matrix, and respectively represent the Euclidean distance between the j th point and the i th point in the base point cloud matrix and the registered point cloud matrix, is the maximum value of the points common to the base point cloud matrix and the registered point cloud matrix.

[0036] Further, in step S31, selecting the nearest points in the base point cloud matrix and the registered point cloud matrix includes:

[0037] S311. Align the coordinate origins of the base point cloud matrix and the registered point cloud matrix, and overlap the base point cloud matrix and the registered point cloud matrix;

[0038] S312. For the point in the base point cloud matrix, select the nearest point from the registered point cloud matrix, satisfying:

[0039] ,

[0040] wherein, represents selecting from the registered point cloud matrix the point that makes take the minimum value as the point in the registered point cloud matrix that is the nearest to the point in the base point cloud matrix, represents the Euclidean distance between the i th point in the base point cloud matrix and the j th point in the registered point cloud matrix.

[0041] Further, after performing point cloud registration on the base point cloud matrix from the lidar and the registered point cloud matrix from the depth camera using the improved point cloud registration algorithm, step S3 further includes:

[0042] S33. Perform stitching processing on the registered multi-frame point cloud matrices, and integrate the multi-frame point cloud matrices into one through the alignment transformation matrix;

[0043] S34. During the integration process, based on the calculated weights, adjust the density of the registered multi-frame point cloud matrices so that the base point cloud matrix from the lidar and the registered point cloud matrix from the depth camera maintain a consistent density in different regions;

[0044] S35. Through iterative optimization, globally adjust the integrated point cloud matrix based on the optimized weight matrix to make the accuracy and consistency of the point cloud map reach the preset threshold overall.

[0045] Further, step S4 includes:

[0046] S41. Data preprocessing: Filter the pose data of the inertial measurement unit to obtain the filtered pose data of the inertial measurement unit.

[0047] S42. State prediction: Predict the current pose data of the inspection robot based on the historical pose data of the inertial measurement unit.

[0048] S43. Kalman gain calculation: Combine the filtered pose data of the inertial measurement unit with the predicted current pose data of the inspection robot to calculate the Kalman gain to determine the weight relationship between the filtered pose data of the inertial measurement unit and the predicted current pose data of the inspection robot.

[0049] S44. State update: Correct the predicted current pose data of the inspection robot according to the Kalman gain to obtain the pose data of the inspection robot.

[0050] Further, step S4 also includes:

[0051] S45. Covariance matrix update: After obtaining the pose data of the inspection robot, update the covariance matrix as the basis for the next pose estimation.

[0052] The present invention also provides a power operation point cloud map construction system based on an inspection robot for implementing the power operation point cloud map construction method based on the inspection robot. The system includes: an inspection robot module: equipped with a variety of sensors, including lidar, depth camera, and inertial measurement unit, for collecting the point cloud matrix of the power operation environment and the pose data of the inspection robot; a data preprocessing module: connected to the inspection robot module, for performing filtering processing and time synchronization on the collected point cloud matrix and the pose data of the inspection robot, and verifying the stability and consistency of the point cloud matrix and the pose data; a pose estimation module: based on the Kalman filtering algorithm, using the data of the inertial measurement unit to estimate the pose of the robot and generating the current pose data of the robot; a point cloud registration module: based on the pose data generated by the pose estimation module, performing the registration operation on the point cloud matrix, and aligning and integrating multiple frames of point cloud matrices into one; a global optimization module: performing global optimization processing on the registered point cloud matrix, adjusting the density of the registered multiple frames of point cloud matrices, so that the basic point cloud matrix from the lidar and the registered point cloud matrix from the depth camera maintain a consistent density in different regions; a refinement processing module: performing refinement processing on the optimized point cloud matrix, removing redundant data and optimizing the structure of the point cloud matrix to generate a high-precision point cloud map; a quality verification and calibration module: performing quality verification and calibration on the generated point cloud map to ensure the accuracy and practicability of the point cloud map.

[0053] Further, the system further includes: a partition management module: after the point cloud map is generated, performing partition management on the point cloud matrix, dividing the map into several regions to improve the calculation efficiency and data management ability; a geometric surface and texture generation module: for the point cloud matrix of each region, generating the corresponding geometric surface and texture information and performing texture mapping processing to enhance the visualization effect of the map; a three-dimensional visualization display module: displaying the finally generated point cloud map on a three-dimensional visualization platform, providing a high-precision three-dimensional view of the power operation environment for the maintenance and management of power facilities. Beneficial effects

[0054] The power operation point cloud map construction method and system based on the inspection robot of the present invention realize the construction of a high-precision three-dimensional point cloud map of the power operation environment by integrating a variety of sensors and advanced data processing technologies. The specific technical effects include:

[0055] 1. High-precision map generation: Through the comprehensive use of sensors such as lidar, depth camera, and IMU, combined with an improved point cloud registration algorithm and Kalman filtering pose estimation, a high-precision three-dimensional point cloud map of the power operation environment can be generated, effectively improving the spatial resolution and detail performance of the map.

[0056] 2. Data Consistency and Optimization: The system ensures the consistency and accuracy of the multi-source point cloud matrix by synchronously calibrating and globally optimizing the collected data. Dynamic weight adjustment and weighted least squares optimization further improve the accuracy of point cloud registration and the overall quality of the map.

[0057] 3. Real-time Performance and Adaptability: The application of the Kalman filtering algorithm in attitude estimation enables the system to process IMU data in real time, quickly respond to environmental changes, and adapt to complex situations in the dynamic power operation environment.

[0058] Enhanced Map Visualization: Through refined processing and three-dimensional visualization, the system not only provides accurate map information but also improves the visual effect of the power operation environment, enabling maintenance and management personnel to understand and operate more intuitively.

[0059] 5. Efficient Data Management: The partition management and geometric surface generation modules of the system improve data processing efficiency, optimize the regional management and visual effect of the map, and facilitate subsequent data analysis and utilization.

[0060] In summary, the method and system of the present invention significantly improve the accuracy, efficiency, and practicality of power operation point cloud map construction, providing effective support for the maintenance and management of power facilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is a schematic flowchart of the method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0062] To deepen the understanding of the present invention, the following will further elaborate on the present invention in combination with embodiments. These embodiments are only used to explain the present invention and do not constitute a limitation on the protection scope of the present invention. Specific Embodiment 1

[0063] According to Figure 1 as shown, this embodiment provides a method for constructing a power operation point cloud map based on an inspection robot. The point cloud map is stored in the form of a point cloud matrix, and the method includes the following steps:

[0064] S1. Obtain multi-source sensor data of the power operation environment. The multi-source sensor data includes a basic point cloud matrix from a lidar, a registered point cloud matrix from a depth camera, and pose data from an inertial measurement unit. The pose data includes position data and attitude data;

[0065] S2. Synchronously calibrate the obtained multi-source sensor data to verify the time consistency of the multi-source sensor data;

[0066] S3. On the premise that the time consistency check passes, use the improved point cloud registration algorithm to perform point cloud registration on the basic point cloud matrix from the lidar and the registered point cloud matrix from the depth camera to obtain the optimized weight matrix;

[0067] S4. Use the Kalman filtering algorithm to perform pose estimation on the inspection robot based on the pose data from the inertial measurement unit and calculate the pose data of the inspection robot;

[0068] S5. Integrate the optimized weight matrix and the pose data of the inspection robot in steps S3 and S4, and use the filtering algorithm to filter and update the integrated point cloud matrix;

[0069] S6. Construct the power operation point cloud map of the power operation environment based on the filtered and updated point cloud matrix.

[0070] Further, the improved point cloud registration algorithm includes:

[0071] S31. Select the closest points in the basic point cloud matrix and the registered point cloud matrix, and calculate the weight of each point. The weight is calculated by the following formula:

[0072] ,

[0073] where, represents the weight of the i th point; is the dynamic adjustment parameter of the i th point, and the initial value is 1; represents the i th point in the basic point cloud matrix, represents the i th point in the registered point cloud matrix with the same serial number as that in the basic point cloud matrix, represents the natural exponential function, and the coordinate parameters of point and point satisfy:

[0074] ,

[0075] where, both the basic point cloud matrix and the registered point cloud matrix are expressed based on the OXYZ orthogonal coordinate system, represents the X-axis coordinate of point , represents the Y-axis coordinate of point , represents the Z-axis coordinate of point , represents the X-axis coordinate of point , represents the X-axis of point ​ Represents the Z-axis coordinate of a point ,

[0076] Represents the Euclidean distance between a point and a point , satisfying:

[0077] ,

[0078] S32. According to the weights of each point, use the weighted least squares method to iteratively optimize the initial weight matrix to obtain an optimized weight matrix, satisfying:

[0079] ,

[0080] wherein, represents the optimized weight matrix; represents the initial weight matrix; the initial weight matrix and the optimized weight matrix each element of which represents the weight value of the corresponding point; the initial values of all elements of the initial weight matrix are 1 and are updated after each iterative optimization;

[0081] represents the maximum value of the number of points in the optimized weight matrix ; represents the iterative optimization function; represents the minimized value; represents the optimal value of the weight that makes the iterative optimization function obtain the minimum value through iterative calculation, as the optimal weight value , and the optimized weight matrix is composed of all the optimal weight values .

[0082] Furthermore, in step S31, the dynamic adjustment parameter i of the th point is adjusted according to the density and dynamics of the base point cloud matrix and the registered point cloud matrix, including: for the i th point, adjusting the dynamic adjustment parameter i of the i th point according to the distribution density of the neighborhood points of the th point, satisfying:

[0083] ,

[0084] wherein, represents the i th point in the base point cloud matrix, Indicates the i th point in the registered point cloud matrix with the same serial number as in the base point cloud matrix, represents the j th point in the base point cloud matrix, indicates the j th point in the registered point cloud matrix with the same serial number as in the base point cloud matrix, and respectively represent the Euclidean distance between the j th point and the i th point in the base point cloud matrix and the registered point cloud matrix, is the maximum value of the points common to the base point cloud matrix and the registered point cloud matrix.

[0085] Furthermore, in step S31, selecting the closest points in the base point cloud matrix and the registered point cloud matrix includes:

[0086] S311. Align the coordinate origins of the base point cloud matrix and the registered point cloud matrix, and overlap the base point cloud matrix and the registered point cloud matrix;

[0087] S312. For the point in the base point cloud matrix, select the closest point from the registered point cloud matrix, satisfying:

[0088] ,

[0089] wherein, represents selecting from the registered point cloud matrix the point that makes take the minimum value , as the closest point in the registered point cloud matrix to the point in the base point cloud matrix, represents the Euclidean distance between the i th point in the base point cloud matrix and the j th point in the registered point cloud matrix.

[0090] Furthermore, after performing point cloud registration on the base point cloud matrix from the lidar and the registered point cloud matrix from the depth camera using the improved point cloud registration algorithm, step S3 further includes::

[0091] S33. Perform stitching processing on the registered multi-frame point cloud matrices, and integrate the multi-frame point cloud matrices into one through the alignment transformation matrix;

[0092] S34. During the integration process, based on the calculated weights, adjust the density of the registered multi-frame point cloud matrices so that the base point cloud matrix from the lidar and the registered point cloud matrix from the depth camera maintain a consistent density in different regions;

[0093] S35. Through iterative optimization, globally adjust the integrated point cloud matrix based on the optimized weight matrix to make the accuracy and consistency of the point cloud map reach the preset threshold as a whole.

[0094] Further, step S4 includes:

[0095] S41. Data preprocessing: Filter the pose data of the inertial measurement unit to obtain the filtered pose data of the inertial measurement unit.

[0096] S42. State prediction: Predict the current pose data of the inspection robot based on the historical pose data of the inertial measurement unit.

[0097] S43. Kalman gain calculation: Combine the filtered pose data of the inertial measurement unit with the predicted current pose data of the inspection robot to calculate the Kalman gain to determine the weight relationship between the filtered pose data of the inertial measurement unit and the predicted current pose data of the inspection robot.

[0098] S44. State update: Correct the predicted current pose data of the inspection robot according to the Kalman gain to obtain the pose data of the inspection robot.

[0099] Further, step S4 also includes:

[0100] S45. Covariance matrix update: After obtaining the pose data of the inspection robot, update the covariance matrix as the basis for the next pose estimation. Specific Embodiment 2

[0101] This embodiment provides a power operation point cloud map construction system based on an inspection robot, which is used to execute the power operation point cloud map construction method based on the inspection robot. The system includes: An inspection robot module: Equipped with a variety of sensors, including lidar, depth camera, and inertial measurement unit, which is used to collect the point cloud matrix of the power operation environment and the pose data of the inspection robot; A data preprocessing module: Connected to the inspection robot module, which is used to perform filtering processing and time synchronization on the collected point cloud matrix and the pose data of the inspection robot, and verify the stability and consistency of the point cloud matrix and the pose data; A pose estimation module: Based on the Kalman filtering algorithm, using the data of the inertial measurement unit to estimate the pose of the robot and generate the current pose data of the robot; A point cloud registration module: Based on the pose data generated by the pose estimation module, perform the registration operation of the point cloud matrix, and align and integrate multiple frames of point cloud matrices into one; A global optimization module: Perform global optimization processing on the registered point cloud matrix to improve the consistency and accuracy of the point cloud matrix and the pose data of the inspection robot and the person; A refinement processing module: Perform refinement processing on the optimized point cloud matrix, remove redundant data and optimize the structure of the point cloud matrix to generate a high-precision point cloud map; A quality verification and calibration module: Perform quality verification and calibration on the generated point cloud map to ensure the accuracy and practicality of the point cloud map.

[0102] Further, the system further includes:

[0103] A partition management module: After the point cloud map is generated, perform partition management on the point cloud matrix, divide the map into several regions to improve the calculation efficiency and data management ability; A geometric surface and texture generation module: For the point cloud matrix of each region, generate the corresponding geometric surface and texture information, and perform texture mapping processing to enhance the visualization effect of the map; A three-dimensional visualization display module: Display the finally generated point cloud map on a three-dimensional visualization platform, provide a high-precision three-dimensional view of the power operation environment for the maintenance and management of power facilities.

[0104] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification only illustrates the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for constructing a point cloud map of electric power operations based on an inspection robot, characterized in that: The electric power operation point cloud map is formed in the form of a point cloud matrix, and the method comprises the following steps: S1. Acquire multi-source sensor data of an electric power operation environment, wherein the multi-source sensor data includes a basic point cloud matrix from a laser radar, a registration point cloud matrix from a depth camera, and position and posture data from an inertial measurement unit, wherein the position and posture data includes position data and attitude data; S2. Synchronize and calibrate the acquired multi-source sensor data to verify the time consistency of the multi-source sensor data; S3. On the premise that the temporal consistency check is passed, the basic point cloud matrix from the lidar and the registration point cloud matrix from the depth camera are registered using the improved point cloud registration algorithm to obtain an optimized weight matrix. S4, using a Kalman filter algorithm, estimating the posture of the inspection robot based on the posture data from the inertial measurement unit, and calculating the posture data of the inspection robot; S5, fusing the optimized weight matrix in step S3 and step S4 with the posture data of the inspection robot, and filtering and updating the fused point cloud matrix using a filtering algorithm; S6. Construct a power operation point cloud map of the power operation environment based on the filtered and updated point cloud matrix.

2. The method for constructing a point cloud map of electric power operations based on an inspection robot according to claim 1, characterized in that: In step S3, the improved point cloud registration algorithm includes: S31, selecting the nearest points in the basic point cloud matrix and the registration point cloud matrix, and calculating the weight of each point, wherein the weight is calculated by the following formula: ， in, Indicates i The weight of each point; For the i The dynamic adjustment parameter of each point, the initial value is 1; Represents the first i Points, Indicates the first i Points, represents the natural exponential function, point Coordinate parameters and points The coordinate parameters satisfy: ， Among them, the basic point cloud matrix and the registration point cloud matrix are both expressed based on the OXYZ orthogonal coordinate system. Indicate point The X-axis coordinate, Indicate point The Y-axis coordinate of Indicate point The Z-axis coordinate of Indicate point The X-axis coordinate, Indicate point The X-axis, Indicate point The Z-axis coordinate of Indicate point and Point The Euclidean distance between them satisfies: ， S32. According to the weight of each point, the initial weight matrix is ​​iteratively optimized using the weighted least squares method to obtain an optimized weight matrix that satisfies: ， in, represents the optimization weight matrix, represents the initial weight matrix, the initial weight matrix and the optimized weight matrix Each element of represents the weight value of the corresponding point, and the initial value of all elements of the initial weight matrix is ​​1, which is updated after each iterative optimization; Represents the optimized weight matrix The maximum number of points in represents an iterative optimization function; represents the minimum value, Represents the iterative optimization function Iteratively calculate the optimal value of the weight that makes it minimum as the optimal weight value , by all the optimal weight values Composition optimization weight matrix .

3. The method for constructing a point cloud map of electric power operations based on an inspection robot according to claim 2, characterized in that: In step S31, i Dynamic adjustment parameters of each point According to the density and dynamic adjustment of the basic point cloud matrix and the registration point cloud matrix, including: i Points, according to i The distribution density of the neighboring points of the point is adjusted i Dynamic adjustment parameters of each point ,satisfy: ， in, Represents the first i Points, Indicates the first i Points, Represents the first j Points, Indicates the first j Points, and Represent the first j Point and i The Euclidean distance between points is The maximum value of the points in common between the base point cloud matrix and the registered point cloud matrix.

4. The method for constructing a point cloud map of electric power operations based on an inspection robot according to claim 3, characterized in that: In step S31, the nearest points in the basic point cloud matrix and the registration point cloud matrix are selected, including: S311, aligning the coordinate origins of the basic point cloud matrix and the registration point cloud matrix, and making the basic point cloud matrix and the registration point cloud matrix overlap; S312, for the points in the basic point cloud matrix , select the nearest point from the registration point cloud matrix ,satisfy: ， in, Indicates that the point cloud matrix is ​​selected so that The point with the smallest value , as the point in the registration point cloud matrix and the base point cloud matrix The nearest point , Represents the first i points and the first in the registration point cloud matrix j The Euclidean distance between points.

5. The method for constructing a point cloud map of electric power operations based on an inspection robot according to claim 2, characterized in that: After performing point cloud registration on the basic point cloud matrix from the laser radar and the registration point cloud matrix from the depth camera using the improved point cloud registration algorithm, step S3 further includes: S33, performing splicing processing on the registered multi-frame point cloud matrices, and integrating the multi-frame point cloud matrices into one by aligning the transformation matrix; S34. During the integration process, based on the calculated weights, the density of the registered multi-frame point cloud matrix is ​​adjusted so that the basic point cloud matrix from the lidar and the registered point cloud matrix from the depth camera maintain a consistent density in different regions; S35. Through iterative optimization, the integrated point cloud matrix is ​​globally adjusted based on the optimized weight matrix, so that the overall accuracy and consistency of the power operation point cloud map reaches a preset threshold.

6. The method for constructing a point cloud map of electric power operations based on an inspection robot according to claim 1, characterized in that: Step S4 includes: S41, data preprocessing: filtering the position and posture data of the inertial measurement unit to obtain filtered position and posture data of the inertial measurement unit; S42, state prediction: predicting the current posture data of the inspection robot based on the historical posture data of the inertial measurement unit; S43, Kalman gain calculation: combining the filtered posture data of the inertial measurement unit with the predicted current posture data of the inspection robot, calculating the Kalman gain to determine the weight relationship between the filtered posture data of the inertial measurement unit and the predicted current posture data of the inspection robot; S44, state update: according to the Kalman gain, the predicted current posture data of the inspection robot is corrected to obtain the posture data of the inspection robot.

7. The method for constructing a point cloud map of electric power operations based on an inspection robot according to claim 6, characterized in that: Step S4 also includes: S45, covariance matrix update: after obtaining the posture data of the inspection robot, the covariance matrix is ​​updated as the basis for the next posture estimation.

8. A system for constructing a point cloud map of electric power operations based on an inspection robot, used to execute a method for constructing a point cloud map of electric power operations based on an inspection robot as claimed in any one of claims 1 to 7, characterized in that: The system comprises: Inspection robot module: equipped with laser radar, depth camera, and inertial measurement unit, used to collect multi-source sensor data of the power operation environment. The multi-source sensor data includes the basic point cloud matrix from the laser radar, the registration point cloud matrix from the depth camera, and the posture data from the inertial measurement unit. The posture data includes position data and attitude data Data preprocessing module: connected to the inspection robot module, used to filter and time-synchronize the collected basic point cloud matrix, registration point cloud matrix and posture data from the inertial measurement unit, and verify the time consistency of multi-source sensor data; Posture estimation module: Use the Kalman filter algorithm to estimate the posture of the inspection robot based on the posture data from the inertial measurement unit and calculate the posture data of the inspection robot; Point cloud registration module: On the premise that the temporal consistency check is passed, the basic point cloud matrix from the lidar and the registration point cloud matrix from the depth camera are registered using the improved point cloud registration algorithm to obtain the optimized weight matrix; Global optimization module: performs global optimization processing on the registered point cloud matrix, adjusts the density of the registered multi-frame point cloud matrix, and makes the basic point cloud matrix from the lidar and the registered point cloud matrix from the depth camera maintain the same density in different areas; Refined processing module: Constructs the power operation point cloud map of the power operation environment based on the filtered and updated point cloud matrix.

9. The power operation point cloud map construction system based on the inspection robot according to claim 8 is characterized in that: The system further comprises: Partition management module: After the power operation point cloud map is generated, the point cloud matrix is ​​partitioned and divided into several areas to improve computing efficiency and data management capabilities; Geometric surface and texture generation module: Generates the corresponding geometric surface and texture information for the point cloud matrix of each area, and performs texture mapping processing to enhance the visualization effect of the point cloud map of the power operation; 3D visualization display module: The final generated power operation point cloud map is displayed on a 3D visualization platform, providing a high-precision 3D view of the power operation environment for maintenance and management of power facilities.

Images