Bolt positioning algorithm applied to power transmission line maintenance

Abstract

The invention discloses a bolt positioning algorithm applied to power transmission line maintenance, and the algorithm comprises the following steps: collecting left and right view angle images of a power transmission line and a bolt through a binocular camera, and recording the motion track of the binocular camera through an IMU; visual data and IMU data are fused by adopting an SLAM technology, a high-precision three-dimensional point cloud map of the power transmission line is constructed, and preliminary screening of bolt areas is carried out; based on the preliminary positioning result of the SLAM, binocular vision and IMU inertial measurement data are combined, and high-precision space correction is carried out through vision inertial fusion; and according to the optimized bolt candidate area, comparing a power transmission line standard model by using a topological structure matching method, and speculating the position of the bolt which may be blocked by dirt or foreign matters. According to the method, binocular vision, IMU data and SLAM technologies are combined, factor graph optimization and Kalman filtering are combined, and topological structure matching and morphological analysis are carried out, so that the positioning precision of the power transmission line bolt can be effectively improved.

Description

Technical Field

[0001] This invention relates to the field of power transmission line maintenance technology, and in particular to a bolt positioning algorithm applied to power transmission line maintenance. Background Technology

[0002] During the routine maintenance and repair of power transmission lines, bolts, as critical fasteners, can cause equipment malfunctions or even serious safety accidents if they become loose, missing, or damaged. Therefore, accurate inspection and positioning of power transmission line bolts is an important task to ensure the safe operation of the power grid. However, since power transmission lines are usually located in complex field environments, traditional manual inspection methods face many challenges, such as high workload, low inspection efficiency, significant susceptibility to environmental factors, and difficulty in meeting the requirements for high precision and high stability.

[0003] In recent years, with the rapid development of robotics, computer vision, and artificial intelligence, automated bolt positioning technology based on visual inspection and intelligent algorithms has gradually become a research hotspot. Currently, some studies use monocular or binocular vision for bolt inspection, but the detection accuracy is low under complex lighting conditions, structural occlusion, or significant changes in viewing angle. Summary of the Invention

[0004] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a bolt positioning algorithm applied to the maintenance of transmission lines, comprising the following steps: Images of the power transmission line and bolts are captured from the left and right perspectives using a binocular camera. At the same time, the movement trajectory of the binocular camera is recorded by an IMU and transmitted to the computing unit. SLAM technology is used to fuse visual data and IMU data to build a high-precision 3D point cloud map of the power transmission line in real time, and the robot's motion trajectory is combined to perform preliminary screening of bolt areas; Based on the preliminary localization results of SLAM, combined with binocular vision and IMU inertial measurement data, high-precision spatial correction is performed through visual-inertial fusion. Based on the optimized bolt candidate region, the topology matching method is used to compare with the standard model of the transmission line to automatically infer the bolt positions that may be obscured by dirt or foreign objects.

[0006] As a preferred embodiment of the bolt positioning algorithm for power transmission line maintenance described in this invention, the acquisition frequency of the binocular camera is synchronized with the inertial measurement frequency of the IMU in a timing sequence.

[0007] As a preferred embodiment of the bolt positioning algorithm for power transmission line maintenance described in this invention, the step of using SLAM technology to fuse visual data and IMU data to construct a high-precision three-dimensional point cloud map of the power transmission line in real time, and combining it with the robot's motion trajectory to perform preliminary screening of bolt areas, includes the following steps: The acquired image data and inertial data are transmitted to the computing unit for fusion processing; In the computing unit, SLAM technology is used to deeply fuse the input visual data and IMU data. Keyframes are generated through front-end feature extraction and matching, and pre-integration optimization is performed in combination with IMU data to construct a high-precision three-dimensional point cloud map of the transmission line. By fusing and analyzing the 3D point cloud map generated by SLAM with the robot's own motion trajectory, and combining the topological information of the power transmission lines in the map, the possible locations of the bolts are initially screened based on the feature point clustering algorithm.

[0008] As a preferred embodiment of the bolt positioning algorithm for power transmission line maintenance described in this invention, the preliminary positioning results based on SLAM, combined with binocular vision and IMU inertial measurement data, are subjected to high-precision spatial correction through visual-inertial fusion, including the following steps: Based on the preliminary localization results of SLAM, feature point detection and matching are performed on the images acquired by the binocular camera, and pre-integration calculation is performed in combination with IMU inertial measurement data to obtain a preliminary estimate of the robot's motion trajectory. Based on this trajectory, preliminary projection correction is performed on the bolt candidate area. Based on the initial projection correction, a factor graph optimization model of vision-inertial fusion is constructed, in which the observation error of binocular vision feature points and the IMU pre-integration error are used as constraint factors, the robot pose and the position of the bolt candidate region are used as state variables, and the optimization solution is obtained by the maximum a posteriori probability estimation algorithm. Based on factor graph optimization, a state estimation algorithm is further used to dynamically correct the position of the bolt candidate region.

[0009] As a preferred embodiment of the bolt positioning algorithm for power transmission line maintenance described in this invention, the state estimation algorithm employs Kalman filtering or nonlinear optimization methods to dynamically correct the bolt candidate region after factor graph optimization, so as to reduce the impact of cumulative error on the final bolt positioning.

[0010] As a preferred embodiment of the bolt positioning algorithm for power transmission line maintenance described in this invention, the dynamic correction formula for the bolt candidate region is:

[0011] in: Indicates time The dynamically corrected candidate bolt region is defined in three-dimensional Euclidean space. ; This represents the position vector of the bolt candidate region after factor graph optimization, which is the estimated value at the previous time step. It represents the correction amount, which is the residual after adjustment by the Kalman gain, and also represents the amount of correction that needs to be made to the current estimate based on the current observation data; Indicates the measured value With predictive models The residuals between the actual measured data and the predicted results represent the deviation between the actual measured data and the predicted results. This represents the current measurement value obtained from binocular vision + IMU observation, i.e., the observed coordinates of the bolt candidate region; For the observation model, it represents the state Predicted observations; The dynamic correction gain matrix, i.e., the Kalman gain, determines the strength of the correction.

[0012] As a preferred embodiment of the bolt positioning algorithm for power transmission line maintenance described in this invention, wherein: the dynamic correction gain matrix The calculation formula is:

[0013] in: Let be the uncertainty covariance matrix of the bolt candidate region, representing the degree of error accumulation after factor map optimization; The prediction error covariance describes the error between the predicted value and the actual measured value. This is the observation matrix, describing the influence of state variables on the measured values; Observation matrix The transpose of a matrix is ​​used to adjust the dimensions of a matrix so that they match the dimensions of the product of other matrices. To observe the noise covariance matrix, describe the uncertainty of binocular vision + IMU measurements; The range of values ​​is The weighting between the optimization results and the observed values ​​is determined by this. and It is the covariance matrix, and its values ​​are positive semi-definite matrices.

[0014] As a preferred embodiment of the bolt positioning algorithm for transmission line maintenance described in this invention, the following steps are included: Based on the optimized bolt candidate region, the bolt positions that may be obscured by dirt or foreign objects are automatically inferred by comparing the topology matching method with the standard model of the transmission line. Based on the optimized candidate bolt locations, the topology information of the transmission line is extracted. Then, the candidate region is matched with the standard model of the transmission line. By comparing the topology and fitting the geometry, the possible locations of bolts are located. Through ensemble morphological analysis, morphological corrosion, dilation, and opening operations are used to identify dirt or foreign objects that may be present in the area around the bolt for bolt locations where there may be obstructions. Based on the variation patterns of these areas, the location of the obstructed bolt is inferred. By comparing the standard model with the possible occlusion area, the results obtained through morphological analysis, combined with the three-dimensional point cloud data of the transmission line and the robot's motion trajectory, are used to correct and optimize the suspected occluded bolts multiple times, improve the accuracy of the bolt restoration position, and generate an accurate bolt position restoration model. The restored bolt positions are compared with the actual observed values, and the accuracy of bolt positioning is gradually improved by using a Kalman filter dynamic correction method.

[0015] As a preferred embodiment of the bolt positioning algorithm for power transmission line maintenance described in this invention, the algorithm further includes a feedback mechanism. This mechanism feeds back the bolt positions that may be obscured by dirt or foreign objects, predicted by topology matching and set morphological analysis, as input to the dynamic correction step to update the state variables. And adjust the observation model This is to achieve continuous improvement in the spatial correction accuracy of the bolt candidate area.

[0016] As a preferred embodiment of the bolt positioning algorithm for power transmission line maintenance described in this invention, when comparing the standard model of the power transmission line using the topology matching method, the standard model of the power transmission line is dynamically updated based on real-time monitoring data to adapt to minor changes in the structure of the power transmission line under different times and environmental conditions.

[0017] The beneficial effects of this invention are: 1. The algorithm provided by this invention combines binocular vision and IMU inertial measurement data, uses SLAM technology to construct a high-precision 3D point cloud map, and performs spatial correction through visual-inertial fusion, enabling precise location of transmission line bolts. Simultaneously, the introduction of factor graph optimization and Kalman filter dynamic correction algorithms effectively reduces accumulated errors, improves the stability and robustness of bolt positioning, and ensures high-precision detection even in complex environments.

[0018] 2. The algorithm provided in this invention compares the standard model of the transmission line using a topology matching method, combines ensemble morphological analysis to automatically infer the position of bolts obscured by dirt or foreign objects, and performs multiple optimizations and corrections using 3D point cloud data and the robot's motion trajectory to ensure the accuracy of bolt repositioning. Even in the presence of obstructions, it can reliably predict and restore the bolt position.

[0019] 3. This invention uses real-time monitoring data to dynamically update the standard model of transmission lines to adapt to minor changes in line structure under different time and environmental conditions. Simultaneously, a feedback mechanism is established, using the results of topology matching and morphological analysis as input to continuously optimize the spatial correction accuracy of bolt candidate areas, thereby improving the long-term adaptability and reliability of the system. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a flowchart of a bolt positioning algorithm for power transmission line maintenance according to the present invention. Detailed Implementation

[0021] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0022] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0023] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0024] Secondly, the present invention is described in detail with reference to the schematic diagrams. When detailing the embodiments of the present invention, for ease of explanation, the cross-sectional views illustrating the device structure may be partially enlarged, not according to the usual scale. Furthermore, the schematic diagrams are merely examples and should not limit the scope of protection of the present invention. In addition, actual fabrication should include three-dimensional spatial dimensions of length, width, and depth.

[0025] Example 1 Reference Figure 1 As an embodiment of the present invention, a bolt positioning algorithm for transmission line maintenance is provided, comprising the following steps: S1: Acquire left and right view images of the power transmission line and bolts using a binocular camera, and simultaneously record the movement trajectory of the binocular camera using an IMU and transmit it to the computing unit.

[0026] Specifically, the acquisition frequency of the binocular cameras is synchronized with the inertial measurement frequency of the IMU to reduce the time deviation between visual and inertial information and improve the fusion accuracy of visual and IMU data. Additionally, this ensures that position calculations are not distorted due to timing errors during SLAM mapping, thus improving the accuracy of spatial correction.

[0027] This step optimizes the synchronization of data acquisition, making SLAM mapping and subsequent localization more accurate.

[0028] S2: SLAM technology is used to fuse visual data and IMU data to build a high-precision 3D point cloud map of the power transmission line in real time, and the bolt area is initially screened by combining the robot's motion trajectory.

[0029] Specifically, SLAM technology is used to fuse visual data and IMU data to construct a high-precision 3D point cloud map of the power transmission line in real time, and the bolt areas are initially screened by combining the robot's motion trajectory, including the following steps: S21: The acquired image data and inertial data are transmitted to the computing unit for fusion processing.

[0030] S22: In the computing unit, SLAM technology is used to deeply fuse the input visual data and IMU data. Keyframes are generated through front-end feature extraction and matching, and pre-integration optimization is performed in combination with IMU data to construct a high-precision 3D point cloud map of the transmission line. The front-end feature extraction and IMU data pre-integration optimization can ensure that SLAM technology can adapt to the complex environment of the transmission line and improve the accuracy of point cloud reconstruction.

[0031] S23: The 3D point cloud map generated by SLAM is fused and analyzed with the robot's own motion trajectory. Combined with the topological information of the power transmission line in the map, the possible areas where the bolts may be located are initially screened based on the feature point clustering algorithm.

[0032] This step provides basic point cloud information for subsequent bolt positioning, and combines it with the robot trajectory to filter candidate bolt regions.

[0033] S3: Based on the preliminary positioning results of SLAM, combined with binocular vision and IMU inertial measurement data, high-precision spatial correction is performed through visual-inertial fusion to improve the accuracy of bolt positioning.

[0034] Specifically, based on the preliminary localization results of SLAM, and combined with binocular vision and IMU inertial measurement data, high-precision spatial correction is performed through visual-inertial fusion, including the following steps: S31: Based on the preliminary localization results of SLAM, feature point detection and matching are performed on the images acquired by the binocular camera, and pre-integration calculation is performed in combination with IMU inertial measurement data to obtain a preliminary estimate of the robot's motion trajectory, and preliminary projection correction is performed on the bolt candidate area based on the trajectory.

[0035] S32: Based on the initial projection correction, a visual-inertial fusion factor graph optimization model is constructed. The observation error of the binocular vision feature points and the IMU pre-integration error are used as constraint factors to improve the SLAM positioning accuracy. The robot pose and the position of the bolt candidate region are used as state variables, and the maximum a posteriori probability estimation algorithm is used for optimization to improve the accuracy of spatial correction.

[0036] S33: Based on factor graph optimization, Kalman filtering or nonlinear optimization methods are further used to dynamically correct the position of the bolt candidate region in order to reduce the impact of cumulative error on the final bolt positioning.

[0037] This step, based on factor graph optimization and maximum a posteriori estimation, improves the accuracy of visual-inertial fusion correction and reduces error accumulation.

[0038] Furthermore, the dynamic correction formula for the bolt candidate region is as follows:

[0039] in: Indicates time The dynamically corrected candidate bolt region is defined in three-dimensional Euclidean space. ; This represents the position vector of the bolt candidate region after factor graph optimization, which is the estimated value at the previous time step. It represents the correction amount, which is the residual after adjustment by the Kalman gain, and also represents the amount of correction that needs to be made to the current estimate based on the current observation data; Indicates the measured value With predictive models The residuals between the actual measured data and the predicted results represent the deviation between the actual measured data and the predicted results. This represents the current measurement value obtained from binocular vision + IMU observation, i.e., the observed coordinates of the bolt candidate region; For the observation model, it represents the state Predicted observations; The dynamic correction gain matrix, i.e., the Kalman gain, determines the strength of the correction, and is calculated using the following formula:

[0040] in: Let be the uncertainty covariance matrix of the bolt candidate region, representing the degree of error accumulation after factor map optimization; The prediction error covariance describes the error between the predicted value and the actual measured value. This is the observation matrix, describing the influence of state variables on the measured values; Observation matrix The transpose of a matrix is ​​used to adjust the dimensions of a matrix so that it is similar to other matrices (such as...). Matching the product dimension of ) To observe the noise covariance matrix, describe the uncertainty of binocular vision + IMU measurements; The range of values ​​is The weighting between the optimization results and the observed values ​​is determined by this. and It is the covariance matrix, and its values ​​are positive semi-definite matrices.

[0041] In summary, the above dynamic correction formula for the bolt candidate region integrates the current observation values. Compared to the state at the previous moment Prediction and utilize the gain term Dynamically adjusting the state estimation and accurately correcting the position of the bolt candidate region after factor map optimization can reduce the impact of cumulative error.

[0042] Example as follows: Known parameters: Initial estimated values ​​(coordinates of the bolts calculated last time):

[0043] Current measurements (bolt coordinates obtained via binocular vision and IMU):

[0044] Predictive Model Calculation: set up Linear motion prediction is used, assuming the previous velocity estimate was... Then the prediction is:

[0045] Kalman gain: set up This represents the trade-off between the current measured value and the predicted value.

[0046] Calculate the measurement residuals:

[0047] Multiply by Kalman gain:

[0048] Calculate the final corrected bolt coordinates:

[0049] Final corrected bolt coordinates Compared to the initial estimate The corrected bolt coordinates are based on dynamic optimization results from historical states and current observations, providing an accurate bolt position estimate.

[0050] S4: Based on the optimized bolt candidate region, the topology matching method is used to compare with the standard model of the transmission line to automatically infer the bolt position that may be blocked by dirt or foreign objects.

[0051] Specifically, based on the optimized bolt candidate region, the topology matching method is used to compare with the standard model of the transmission line to automatically infer the location of bolts that may be obscured by dirt or foreign objects. This includes the following steps: S41: Based on the optimized candidate bolt locations, the topology information of the transmission line is extracted, including the geometry and relative positions of components such as towers, conductors, joints, and bolts. Then, this candidate region is matched with the standard model of the transmission line. Through topology comparison and geometry fitting, the possible locations of bolts are identified. Furthermore, when using the topology matching method to compare with the standard model of the transmission line, the standard model is dynamically updated based on real-time monitoring data to adapt to minor changes in the transmission line structure under different time and environmental conditions, thereby improving the accuracy and adaptability of the matching.

[0052] S42: Through ensemble morphological analysis, morphological corrosion, dilation, and opening operations are used to identify dirt or foreign objects that may be present in the area around the bolt for bolt locations where there may be obstructions. Based on the variation patterns of these areas, the location of the obstructed bolt is inferred.

[0053] S43: By comparing the standard model with the possible occlusion area, the results obtained through morphological analysis, combined with the three-dimensional point cloud data of the transmission line and the robot's motion trajectory, are used to make multiple corrections and optimizations to the suspected occlusion bolts, improve the accuracy of the bolt restoration position, and generate an accurate bolt position restoration model.

[0054] S44: Compare the restored bolt position with the actual observed value, and gradually improve the bolt positioning accuracy through Kalman filtering dynamic correction method.

[0055] Furthermore, the algorithm includes a feedback mechanism that feeds back the predicted bolt locations that may be obscured by dirt or foreign objects, based on topological matching and set morphological analysis, to the dynamic correction step to update the state variables. And adjust the observation model This is to achieve continuous improvement in the spatial correction accuracy of the bolt candidate area.

[0056] In summary, the algorithm provided by this invention combines binocular vision and IMU inertial measurement data, employs SLAM technology to construct a high-precision 3D point cloud map, and performs spatial correction through visual-inertial fusion, enabling precise location of transmission line bolts. Simultaneously, the introduction of factor graph optimization and Kalman filter dynamic correction algorithms effectively reduces accumulated errors, improves the stability and robustness of bolt positioning, and ensures high-precision detection even in complex environments. The algorithm compares the standard transmission line model using a topology matching method, combines ensemble morphological analysis to automatically infer bolt positions obscured by dirt or foreign objects, and utilizes 3D point cloud data and robot motion trajectories for multiple optimization corrections to ensure accurate bolt recovery and positioning. Even in the presence of obstructions, it can reliably predict and recover bolt positions. This invention uses real-time monitoring data to dynamically update the standard transmission line model to adapt to minor changes in the line structure under different time and environmental conditions. Furthermore, a feedback mechanism is established, using the results of topology matching and morphological analysis as input to continuously optimize the spatial correction accuracy of the bolt candidate region, thereby improving the long-term adaptability and reliability of the system.

[0057] Example 2 is an embodiment of the present invention, which provides a bolt positioning algorithm for power transmission line maintenance. In order to verify the beneficial effects of the present invention, simulation experiments are conducted to analyze the following two aspects: positioning accuracy and recovery capability under obstruction conditions.

[0058] The experiment was conducted in a virtual simulation environment, simulating power transmission line scenarios under different environmental conditions, including normal and obstructed environments. Data acquisition employed a binocular vision system and an IMU (Inertial Measurement Unit), combined with SLAM (Simultaneous Localization and Mapping) technology to construct a high-precision 3D point cloud map.

[0059] The experiment was divided into two parts: verification of positional accuracy and verification of recovery capability under occlusion conditions.

[0060] When verifying the location accuracy, transmission line models of different complexities (simple, medium, and complex) were selected. The algorithm of this invention was compared with the traditional visual SLAM-based method. Then, the average positioning error under different environments was calculated. Table 1 below shows the experimental results.

[0061] Table 1

[0062] Conclusion: The algorithm of this invention, which combines IMU data with visual-inertial fusion, can significantly reduce positioning errors, especially in complex environments where the improvement is more pronounced.

[0063] When verifying the recovery capability under occlusion conditions, different degrees of occlusion (mild, moderate, and severe) were artificially created in the simulation environment. The recovery success rate of the traditional method and the algorithm of this invention were then compared. Table 2 below shows the experimental results.

[0064] Table 2

[0065] Conclusion: This invention, combining topological matching and set morphological analysis, improves the bolt recovery rate under occlusion conditions, especially in heavily occluded environments, where the recovery success rate is significantly enhanced. Furthermore, traditional methods struggle to recover bolt positions in highly occluded environments, while this invention can accurately predict and correct the location of occluded bolts.

[0066] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A bolt positioning algorithm applied to the maintenance of a power transmission line, characterized in that, The method comprises the following steps: Collecting left and right view images of the power transmission line and the bolt through a binocular camera, recording the motion trajectory of the binocular camera through an IMU, and transmitting to a computing unit; Fusing visual data and IMU data by using SLAM technology, constructing a high-precision three-dimensional point cloud map of the power transmission line in real time, and preliminarily screening the bolt area in combination with the motion trajectory of the robot; Based on the preliminary positioning result of SLAM, combining binocular vision and IMU inertial measurement data, high-precision spatial correction is performed through visual-inertial fusion. According to the optimized bolt candidate area, the topological structure matching method is used to compare the standard model of the power transmission line to automatically infer the position of the bolt that may be blocked by dirt or foreign matter.

2. The bolt positioning algorithm for transmission line maintenance of claim 1, wherein: The acquisition frequency of the binocular camera and the inertial measurement frequency of the IMU are time-synchronized.

3. The bolt positioning algorithm for transmission line maintenance of claim 1, wherein: The step of fusing visual data and IMU data by using SLAM technology, constructing a high-precision three-dimensional point cloud map of the power transmission line in real time, and preliminarily screening the bolt area in combination with the motion trajectory of the robot comprises the following steps: Transmit the acquired image data and inertial data to the computing unit for fusion processing; In the computing unit, the input visual data and IMU data are deeply fused by using SLAM technology, key frames are generated through front-end feature extraction and matching, and pre-integration optimization is performed in combination with IMU data, so as to construct a high-precision three-dimensional point cloud map of the power transmission line; Through the fusion analysis of the three-dimensional point cloud map generated by SLAM and the motion trajectory of the robot itself, the topological structure information of the power transmission line in the map is combined, and the feature point clustering algorithm is used to preliminarily screen the area where the bolt may be located.

4. The bolt positioning algorithm for transmission line maintenance of claim 1, wherein: The step of performing high-precision spatial correction through visual-inertial fusion based on the preliminary positioning result of SLAM in combination with binocular vision and IMU inertial measurement data comprises the following steps: Based on the preliminary positioning result of SLAM, feature point detection and matching are performed on the images collected by the binocular camera, pre-integration calculation is performed in combination with IMU inertial measurement data to obtain a preliminary estimate of the motion trajectory of the robot, and the bolt candidate area is preliminarily projected and corrected based on the trajectory; Based on the preliminary projection correction, a visual-inertial fusion factor graph optimization model is constructed, wherein the observation error of the binocular vision feature points and the pre-integration error of the IMU are used as constraint factors, the position of the robot pose and the bolt candidate area are used as state variables, and the maximum a posteriori probability estimation algorithm is used for optimization and solving; Based on the factor graph optimization, a state estimation algorithm is further used to dynamically correct the position of the bolt candidate area.

5. The bolt positioning algorithm for transmission line maintenance of claim 4, wherein: The state estimation algorithm uses Kalman filtering or a nonlinear optimization method to dynamically correct the bolt candidate area after factor graph optimization, so as to reduce the influence of cumulative error on the final bolt positioning.

6. The bolt positioning algorithm for transmission line maintenance of claim 5, wherein: The dynamic correction formula of the bolt candidate area is: ； wherein: representing a time the bolt candidate region position after dynamic correction, the domain being a three-dimensional Euclidean space ; represents the position vector of the bolt candidate region after the factor graph optimization, i.e., the estimation value at the previous time. represents the correction amount, i.e. the residual adjusted by the Kalman gain, and also represents the amount of correction to be made to the current estimate based on the current observation data; Indicates the measured value With predictive models The residuals between the actual measured data and the predicted results represent the deviation between the actual measured data and the predicted results. represents the measurement value at the current time obtained by binocular vision + IMU observation, that is, the observation coordinates of the bolt candidate region; For the observation model, the predicted observation value is represented by the state xt The strength of the correction is determined by the dynamic correction of the gain matrix, i.e. the Kalman gain.

7. The bolt positioning algorithm for transmission line maintenance of claim 5, wherein: The dynamic correction gain matrix The calculation formula is: ； wherein: is the uncertainty covariance matrix of the bolt candidate region, representing the cumulative degree of error after factor graph optimization; is the prediction error covariance, which describes the error between the predicted value and the actual measured value; is the observation matrix, describing the influence of the state variable on the measured value; is the transpose of the observation matrix , the transpose operation is used to adjust the dimension of the matrix, so that the product dimension of it and other matrices matches; is the observation noise covariance matrix, describing the uncertainty of binocular vision + IMU measurement; The value range between determines the weight distribution between the optimization result and the observation value; and is the covariance matrix, and the value range is a semi-definite matrix.

8. The bolt positioning algorithm for transmission line maintenance of claim 5, wherein: According to the optimized bolt candidate area, the topological structure matching method is used to compare the standard model of the power transmission line to automatically infer the position of the bolt that may be blocked by dirt or foreign matter. According to the optimized bolt candidate region position, the topological structure information of the power transmission line is extracted, and then the candidate region is matched with the standard model of the power transmission line. Through the comparison of the topological structure and the fitting of the geometric shape, the position of the possible bolt is located; Through set morphological analysis, for the bolt position that may exist occlusion, morphological erosion, dilation and opening operation are used to identify the dirt or foreign matter that may exist in the surrounding area of the bolt, and according to the change rule of these areas, the occluded bolt position is inferred; By comparing the standard model with the possible occlusion area, combining the three-dimensional point cloud data of the power transmission line and the motion trajectory of the robot, the suspected occluded bolt is corrected and optimized for multiple times, and a precise bolt position recovery model is generated; The restored bolt position is compared with the actual observation value, and the Kalman filtering dynamic correction method is used to gradually improve the accuracy of bolt positioning.

9. The bolt positioning algorithm for transmission line maintenance of claim 8, wherein: The algorithm further comprises a feedback mechanism that feeds back the predicted bolt locations that are likely to be obscured by dirt or foreign matter based on the topological structure matching and set morphological analysis as input to the dynamic revision step to update the state variables and adjust the observation model to achieve continuous improvement of the spatial correction accuracy of the bolt candidate regions.

10. The bolt positioning algorithm for transmission line maintenance of claim 8, wherein: When the topological structure matching method is used to compare the standard model of the power transmission line, the standard model of the power transmission line is dynamically updated according to the real-time monitoring data to adapt to the slight changes of the power transmission line structure under different time and environmental conditions.

Images