Power transmission line fitting method for LiDAR robot point cloud power inspection

Abstract

According to the power transmission line fitting method for the LiDAR robot point cloud power inspection, the LiDAR system is carried on the robot, data processing is carried out on the scanning precision of the LiDAR system and the three-dimensional point cloud, and a new power transmission line inspection mode which is high in economic benefit and adapts to future development is explored. Firstly, data measured and collected by the LiDAR robot are analyzed, the geometric positioning principle of the data and the conversion relation of coordinate systems of all components are established, then the working process of an airborne laser radar measurement system is designed, and a precision analysis and calibration method for point cloud data of the LiDAR robot is provided. Component models of ranging errors, instantaneous scanning angle errors, system placement errors, system placement errors and attitude angle errors are subjected to quantitative analysis under the working background of a line patrol robot, and the relative precision of a point cloud model is improved by performing on-flight calibration on an airborne laser radar measurement system. The point cloud data of the LiDAR robot is used for fitting a power line equation, and safe, rapid, accurate and economical power corridor inspection is achieved.

Description

Technical Field

[0001] The present application relates to a LiDAR robot point cloud transmission line fitting method, and in particular to a LiDAR robot point cloud power inspection transmission line fitting method, belonging to the technical field of LiDAR point cloud transmission line fitting. Background Art

[0002] Transmission lines are the core components of the power grid system. However, since transmission lines are usually laid in the open in the wild with complex terrain and harsh conditions, they are often affected by strong winds, rain, snow, lightning, sand, sunlight, birds and the surrounding environment, which causes aging and corrosion of power lines and components. Therefore, for large-scale power grids, regular inspections of transmission lines to ensure the safety and stability of power lines are an important task in current power grid maintenance.

[0003] Manual inspection is a traditional inspection method, which requires maintenance personnel to cross mountains and work in the field for a long time. The working environment in the field is difficult, and the inspection time is long, and the work efficiency is relatively low due to running around in the field all year round. In addition, once a special situation occurs, such as entering the wilderness, or the transmission line crossing a large river, the inspection difficulty is greatly increased, which brings pressure to the inspection workers. Therefore, considering that manual inspection is subject to external environment and human restrictions, it is easy to cause the safety hazards of the transmission line to be unable to be detected in time, and eventually cause disasters. New methods should be adopted to inspect the transmission line to reduce the safety hazards in the transmission line. The traditional manual inspection method can no longer meet the development and safe operation needs of the modern power grid.

[0004] In order to maintain a safe distance from the line tower, aerial inspections cannot inspect the line towers at close range. At the same time, they are greatly affected by weather factors and cannot inspect in severe conditions such as heavy fog and strong winds. In addition, helicopter inspections cover many issues such as personnel, navigation, ground maintenance, fuel, communications, weather, and repairs. Drone inspections are small and flexible, and can inspect the lines at close range with high accuracy, but their payload is limited and they cannot carry too many inspection equipment. At the same time, they are restricted by battery factors and have a short battery life.

[0005] Robot inspection uses intelligent robots as carriers, and the phase line or ground line of the transmission line as the operation path. It carries detection equipment to inspect the transmission line corridor. Because it is close to the transmission line, the detection accuracy is high. In addition to routine detection of the line and equipment, it can also perform fault repairs on the power line through different additional structures, replace the spacer rods, and trim the vegetation near the power line. It has the advantages of powerful inspection function, low maintenance cost, and long battery life.

[0006] The laser detection and ranging system LiDAR can accurately measure the distance from the detector to the object or the surface. By processing the data generated during the ranging process, the digital ground model DEM can be obtained. Through the point cloud data, the three-dimensional space model of the measured object can be constructed. It is conducive to collecting the coordinates along the power grid and obtaining a high-precision point cloud model of the transmission corridor, including the actual geometric data of the transmission line, towers, vegetation, buildings, etc., and recording these point cloud data in a short time, so that the distance between the wire and the ground can be further automatically measured, or the distance between the wire and the adjacent wire can be measured. Through data analysis of the spacing distance between wires, the vertical curvature of the transmission line or the spacing between power towers, etc., a series of data is generated to determine whether the tension of the power optical cable meets the requirements, so as to take measures in advance to prevent damage to the power optical cable. The data obtained by LiDAR measurement can obtain the power grid safety status monitoring and management data, reducing the cost of inspection and safety status management.

[0007] However, the existing power inspection transmission line fitting method has many problems and key technical difficulties to be solved, including:

[0008] (1) The existing technology of manual inspection of power transmission lines requires maintenance personnel to cross mountains and work in the field for a long time. It is hard work and the inspection time is long, so the work efficiency is relatively low. Once entering the wilderness and crossing large rivers, the inspection difficulty is greatly increased, which brings huge pressure to the inspection workers, and it is easy to cause the safety hazards of the power transmission lines to fail to be detected in time, which eventually leads to disasters. In order to maintain a safe distance from the tower, aerial inspection cannot inspect the tower at close range, and it is also impossible to inspect in severe conditions such as heavy fog and strong wind. In addition, helicopter inspection covers many issues such as personnel, navigation, ground maintenance, fuel, communication, weather, and repair. UAV inspection is small and flexible, but its payload is limited and it cannot carry too many detection equipment. At the same time, it is restricted by battery factors and has a short battery life. The existing technology lacks the application of combining power robot inspection and LiDAR system point cloud data collection for power line inspection, which cannot improve data collection and analysis capabilities. The power grid detection speed is slow, the accuracy is low, and the cost is high.

[0009] (2) The introduction and configuration of LiDAR system to the line inspection robot faces a series of difficulties. The positioning rules of LiDAR system need to be defined and tested on the power line inspection robot. The point cloud measurement accuracy of transmission line inspection needs to be further improved. The error of LiDAR robot power transmission line inspection needs to be analyzed and reduced. It is impossible to extract the power line equation using point cloud data. Since the traditional airborne laser radar measurement system is a LiDAR system mounted on an aircraft, and the laser radar measurement system is mounted on the inspection robot, its geometric positioning and the conversion relationship of the coordinate system of each component need to be recalibrated. The workflow of the airborne laser radar measurement system will also change. The error sources of LiDAR robot power inspection are many. It is necessary to quantitatively analyze the ranging error, instantaneous scanning angle error, system placement error (including placement angle error and eccentricity measurement error), system placement error, and attitude angle error component model in the context of the line inspection robot. It is necessary to calibrate the airborne laser radar measurement system in flight to improve the relative accuracy of the point cloud model and meet the needs of robot point cloud power transmission line inspection and fitting.

[0010] (3) The introduction of the LiDAR system and its configuration on the line inspection robot poses new challenges to the positioning, point cloud measurement accuracy and error analysis and correction of the LiDAR system. At the same time, due to the limitations of the overhead line catenary equation in practical applications, the existing technology lacks a method for extracting and fitting the transmission line for LiDAR robot point cloud power inspection. It is impossible to fit the overhead line catenary equation based on the LiDAR robot system, and it is impossible to combine the transmission line point cloud model generated by the airborne laser radar measurement equipment. It is impossible to use the elevation projection method for filtering, the Canny operator edge detection and the improved Hough line detection method to extract the power line, and it is impossible to use the point cloud data of the power line to fit the power line equation. As a result, the accuracy of generating the power line equation using the point cloud model is not high, and the efficiency is reduced, which ultimately leads to the inability to safely, quickly, accurately and economically inspect the power corridor. Summary of the invention

[0011] This application installs the LiDAR system on a robot to inspect the power corridor, and uses its scanning accuracy and three-dimensional point cloud for data processing to explore a new, economically efficient and future-friendly way of power transmission line inspection. First, the data collected by the LiDAR robot is analyzed, and its geometric positioning principle and the conversion relationship between the coordinate systems of each component are established. Then, the workflow of the airborne laser radar measurement system is designed, and the accuracy analysis and calibration method of the LiDAR robot point cloud data is proposed. The ranging error, instantaneous scanning angle error, system placement error (including placement angle error and eccentricity measurement error), system placement error, and attitude angle error component model are quantitatively analyzed in the working background of the line patrol robot. By performing in-flight calibration of the airborne laser radar measurement system, the relative accuracy of the point cloud model is improved. The power line equation is fitted using the LiDAR robot point cloud data to achieve safe, fast, accurate, and economical inspection of the power corridor.

[0012] In order to achieve the above technical effects, the technical solutions adopted in this application are as follows:

[0013] The transmission line fitting method of LiDAR robot point cloud power inspection introduces the LiDAR system and configures it on the line inspection robot. Based on the positioning of the LiDAR system by the intelligent robot, the point cloud measurement accuracy and error analysis correction, the point cloud data is used to extract the power line equation;

[0014] S1: Build a LiDAR robot power inspection platform mounted on an intelligent robot, and form a complete workflow for data collection and generation of a point cloud model of a power transmission line using an airborne LiDAR measurement system;

[0015] S2: Establish multiple error models for configuring the LiDAR system on the line patrol robot, quantitatively analyze the impact of multiple errors on the final laser foot point coordinate accuracy, significantly reduce the laser foot point coordinate accuracy based on the comprehensive error, and integrate multiple error sources to perform integrated calibration of the system;

[0016] S3: Based on the inspection features of the intelligent robot for power transmission lines, the airborne laser radar measurement equipment is calibrated in flight. By correcting the roll angle and heading angle, the point cloud model image generated by the measurement is made clear, thereby improving the relative accuracy.

[0017] S4: Establish a method for fitting the catenary equation of overhead lines based on the LiDAR robot system. Combined with the point cloud model of the transmission line generated by the airborne laser radar measurement equipment, a power line equation fitting framework is constructed. The elevation distribution features and plane projection features of the LiDAR power line scanning data are extracted. The elevation projection method is used for filtering, elevation projection and resampling, Canny operator edge detection and improved Hough line detection methods are used to extract power lines. After clustering of adjacent pixels, perception grouping of lines, and line segment detection based on free probability transformation, the power line equation is fitted using the point cloud data of the power line. It is concluded that generating the power line equation using the point cloud model is more accurate and faster.

[0018] Preferably, the power line equation fitting framework: the shape of the power line is calculated by using the catenary equation of the overhead power line, and the integral form of the catenary equation of the overhead power line is:

[0019]

[0020] where σ 0 is the horizontal stress of the overhead line, γ is the specific load, C 1 , C 2 is a constant whose value is related to the origin of the coordinate system;

[0021] The equation of overhead wire with unequal height suspension points is used as the mathematical model of overhead ground wire. The height difference between the two suspension points A and B is the height difference h, and the angle between the two suspension points and the horizontal plane is the height difference angle Φ. Point A is taken as the origin of the coordinate system:

[0022]

[0023] in, It is expressed as the length of the catenary in the overhead line with equal height suspension points;

[0024] The catenary shape of the overhead line As long as the ratio is the same, the final suspension curve of the overhead line is the same, and as long as The final suspension curve of the overhead line is certain, where y is the only influencing factor;

[0025] The power lines are scanned using a laser radar measurement system to generate a three-dimensional point cloud model of the power lines. The point cloud model is used to accurately determine the spatial three-dimensional coordinates of any point on the power lines and the specific shape of the power lines. A power line equation consistent with the actual situation is generated and fed back to the overhead line catenary equation, thereby providing accurate data parameters for power line research.

[0026] Preferably, the elevation distribution feature extraction: based on the power line structure, within a certain range, ensure that there is a difference between the power line and the ground elevation, and use the elevation difference to preliminarily distinguish the ground points and the power line;

[0027] Among the measured data points, the vast majority are concentrated in areas with smaller elevation values, which are ground points. Only a small number of data points are concentrated in areas with larger elevation values, which are power line points. Scattered between ground points and power line points are tower points and other ground features. Within a certain interval, power lines have basically the same elevation distribution or are similarly on the same horizontal plane.

[0028] Preferably, plane projection feature extraction: the point cloud data obtained after filtering includes the following: power line point cloud data, pole tower point cloud data, and some ground object point cloud data that are not filtered out. In these point cloud data, the ground objects that are not filtered out may be close to the power lines, which is less than the safe distance. The projection of the point cloud data on the horizontal plane after filtering shows that: the power lines appear as parallel lines spliced ​​together by discrete discontinuity points, which are extracted by linear fitting to obtain power line extraction data; pole towers and possible dangerous ground objects are both irregular closed figures, which are separated by mathematical morphology methods. In addition, possible dangerous ground objects are vegetation or buildings, and their projected area is smaller than that of pole towers. An area critical value is set to separate the two and distinguish between pole towers and possible dangerous ground objects.

[0029] Preferably, a power line fitting algorithm based on elevation projection: initially, the filtering analysis is completed by using the elevation critical value segmentation method. First, some ground points are removed, and the remaining data obtained are power line point cloud data, pole tower point cloud data, and some ground object point cloud data that are not filtered out; then, the above data are converted into a two-dimensional elevation value projection through the elevation projection method and the resampling method, and the power lines are extracted and fitted using the linear fitting method to obtain power line related data; finally, the obtained image is processed to facilitate the separation of pole towers and possible dangerous objects.

[0030] Preferably, the filtering of the power line fitting algorithm of the elevation projection is as follows: according to the distribution characteristics of the data, the power line point cloud data is obtained, and then the ground points are removed by using the elevation critical value segmentation method to reduce the data and improve the fitting degree and fitting efficiency. The optimal iterative critical value method is used to achieve filtering. The specific calculation steps are as follows:

[0031] Step 1: Select the average elevation T 0 As the initial critical value of this calculation, that is, T k =T 0 ;

[0032] Step 2: According to the critical value T k Then the data is divided into two subsets, and the average elevation of the two subsets is calculated in turn, which are T A and T B , assuming T A <Tk , T B >T k ;

[0033] Step 3: Get the critical value T according to formula 3 k+1 :

[0034]

[0035] Step 4: T k+1 Instead of T k ;

[0036] Step 5: Repeat the above process until T k Convergence, numerically T k+1 =T k , then T k is the selected segmentation threshold T;

[0037] Step 6: Based on the final segmentation threshold T obtained, the data is divided into two parts.

[0038] Preferably, elevation projection and resampling: using point cloud projection and resampling methods to obtain elevation value images, calculating the coverage and elevation range of the data set on the XOY plane; mapping non-ground points to the XOY plane, and calculating the projection sampling distance d according to the width W of the elevation value image by formula 4:

[0039]

[0040] The plane coordinates (x, y) of a point are converted into the corresponding grid row and column numbers (r, c) by equation 5:

[0041]

[0042] Resampling is applied point by point. If there is no laser point in the grid point, the gray value is zero. If there is a laser point in the grid point, the maximum elevation value is normalized to the gray range of [a, b] using the formula and regarded as the gray value gray:

[0043]

[0044] Finally, the elevation value image is obtained, which realizes the transformation of point cloud data from three-dimensional space to two-dimensional plane.

[0045] Preferably, clustering of adjacent pixels: after filtering, segmenting, thinning, edge processing, etc. of the image actually acquired, clustering of adjacent pixels in the image is performed to reduce interference of short lines formed by noise on detection. The specific steps are as follows:

[0046] Step 1: First select a point P in the binary image 0, then determine whether the point is 1, that is, P 0 Is it a feature pixel? If the obtained P 0 ≠1, then select a point in the binary image again. If P 0 =1, then enter the next stage;

[0047] Step 2: Record the position of the obtained feature pixel point as P, put it into one end of the feature point set S+P→S, and set the pixel value of the point to P=0;

[0048] Step 3: Judge all the neighborhoods of point P to see if they are feature points. If the final judgment result is a feature point, repeat the second step. If the final judgment result is not a feature point, go to the next step.

[0049] Step 4: Re-align P 0 Search the data in the neighborhood of the point to determine whether these data have connected feature pixels. If the judgment result is yes, execute the fifth step. If the judgment result is no, terminate the execution.

[0050] Step 5: Put the feature points into the other end of the set P+S→S, and the rest of the operations are the same as step 2;

[0051] Step 6: Continue to determine whether these data are feature points. If the determination result is yes, execute step 5. If the determination result is no, terminate the execution.

[0052] After the above processing, the pixels are aggregated. In addition, the independent pixels are removed and the pixels are regarded as similar to straight lines. This will further determine the detection object and separate the parallel lines with close intervals.

[0053] Preferably, the perceptual grouping of straight lines: further subdividing the polyline will make the resulting polyline closer to a straight line, (x 1 ,y 1 )、(x 2 ,y 2 ) represents the coordinates of the two end points of the short line, and B is the coordinates of the short line passing through (x 1 ,y 1 )、(x 2 ,y 2 ) of the straight line, (x 3 ,y 3 ) represents the coordinates of the point where line segment A deviates from line B by a distance d, and d is calculated by formula 7:

[0054] d=[(y 3 -y 1 )(x 2 -x 1 )-(x3 -x 1 )(y 2 -y 1 )] / L Formula 7

[0055] Where: make:

[0056] R=L / maxd Formula 8

[0057] The ratio R is only related to the line itself, and has nothing to do with the size of the image obtained, and does not change with the size of the image. The point with the largest offset distance is used for binary segmentation to obtain the ratio R i (i=1, 2), compare it with the ratio of the original line segment. If the ratio R i If the value of is larger, the original line segment is replaced by two sub-line segments; otherwise, the original line segment is retained, so that the original line segment is closer to a straight line and no subsequent processing is required;

[0058] Through calculation and analysis, each line segment will be closer to a straight line, determining the detected straight line and improving the detection accuracy.

[0059] Preferably, the straight line segment detection based on the free probability transformation is as follows: the approximate straight line segments are detected in sequence, and two points (x p ,y p )、(x q ,y q ), use two points to generate a straight line, and the generated straight line variable is obtained by the following formula:

[0060]

[0061] ρ=x p cosθ+y p sinθ or ρ = x q cosθ+y q sinθ Equation 10

[0062] The use of clustering and perceptual grouping can better ensure effectiveness, so when performing Hough transform on the above image, it only needs to reach a very small critical value. The above analytical calculation results are as follows:

[0063] Step 1: Get all the pixels d i (x i ,y i ) is regarded as a set D;

[0064] Step 2: Select d from the above set D. p , d q Calculate whether they meet the requirements of a point pair. If the two selected d p , dq If the point does not meet the point pair requirement, reselect d p , d q ;

[0065] Step 3: Find d p , d q The corresponding point (θ, ρ) is accumulated to 1 at the corresponding position;

[0066] Step 4: If the value of a certain point in the parameter space reaches the critical value, the above calculation is stopped, and the straight line corresponding to this point is the desired straight line;

[0067] Step 5: Use OpenCV software to process the elevation projection. Through software calculation, the projections of most power lines are detected.

[0068] Sort the detected straight lines by their lengths p in the polar coordinate system, and manually select the power line matching line that best meets the needs by comparing the straight lines corresponding to different p values;

[0069] Finally, the matching power line point cloud data is separated to obtain the three-dimensional coordinates of the point cloud, and the point cloud coordinates are input into the MATLAB software to display its spatial position and form.

[0070] Compared with the prior art, the innovations and advantages of this application are:

[0071] (1) This application installs a LiDAR system on a robot to inspect the power corridor, and uses its scanning accuracy and three-dimensional point cloud for data processing, exploring a new, economical and future-friendly way of power transmission line inspection. First, the data collected by the LiDAR robot is analyzed, and its geometric positioning principle and the conversion relationship between the coordinate systems of each component are established. Then, the workflow of the airborne laser radar measurement system is designed, and the accuracy analysis and calibration method of the LiDAR robot point cloud data is proposed. The ranging error, instantaneous scanning angle error, system placement error (including placement angle error and eccentricity measurement error), system placement error, and attitude angle error component model are quantitatively analyzed in the working background of the line inspection robot. By performing in-flight calibration of the airborne laser radar measurement system, the relative accuracy of the point cloud model is improved. The power line equation is fitted using the LiDAR robot point cloud data to achieve safe, fast, accurate and economical inspection of the power corridor.

[0072] (2) This application introduces the LiDAR system and configures it on the line patrol robot. Based on the positioning of the LiDAR system by the intelligent robot, the point cloud measurement accuracy and error analysis correction, the point cloud data is used to extract the power line equation; a laser radar measurement system mounted on the intelligent robot is constructed, and a complete set of workflows for using the airborne laser radar measurement system to collect data and generate the point cloud model of the transmission line is formed. A variety of error models are constructed in detail, and combined with the characteristics of the high-voltage transmission line patrol robot inspection, the influence of various errors on the final laser foot point coordinate accuracy is quantitatively analyzed. It is found that the comprehensive error will significantly reduce the laser foot point coordinate accuracy, but the impact on it is not a simple linear superposition of various errors. Its influence mechanism is very complex, and it is necessary to integrate multiple error sources before the system can be integrated and calibrated. In view of the inspection characteristics of the high-voltage transmission line patrol robot, the airborne laser radar measurement equipment is calibrated in flight. By correcting the roll angle and heading angle, the point cloud model image generated by the measurement is finally made clear and the relative accuracy is improved. Due to the limitations of the overhead line catenary equation in practical applications, a fitting method for the overhead line catenary equation based on the LiDAR robot system is proposed. The point cloud model is used to generate the power line equation more accurately and quickly.

[0073] (3) This application proposes a method for fitting the catenary equation of overhead lines based on a LiDAR robot system, which solves the limitations of the catenary equation of overhead lines in practical applications. It combines the point cloud model of the power line generated by the airborne laser radar measurement equipment, uses the elevation projection method to filter, the Canny operator edge detection and the improved Hough line detection method to extract the power lines, and uses the point cloud data of the power lines to fit the power line equation, and draws a conclusion that it is more accurate and faster to generate the power line equation using the point cloud model. This method has the outstanding advantages of safe driving along the line (track), the robot reaches wherever the line goes, strong environmental adaptability, easy online power supply, high line patrol efficiency, good close-range patrol effect, unmanned or few-person operation, and low cost. In addition, as a three-dimensional measurement system, the LiDAR system can quickly obtain three-dimensional point cloud data of the line environment, realize the safety status monitoring and management of the power grid, and greatly reduce the cost of patrol and safety status management. Mounting the LiDAR system on an intelligent robot to inspect transmission lines reduces the cost of inspection and safety status management. It is a new, cost-effective way of inspecting transmission lines that is adaptable to future development. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 This is a schematic diagram of the catenary of an overhead line with unequal height suspension points.

[0075] Figure 2 It is a schematic diagram of the elevation of power line scanning data.

[0076] Figure 3It is the elevation distribution histogram of power line scanning data.

[0077] Figure 4 It is a cross-sectional diagram of the ground point filtering result.

[0078] Figure 5 It is the flow chart of the clustering algorithm of adjacent pixels.

[0079] Figure 6 This is a schematic diagram of calculating the maximum offset distance of a straight line.

[0080] Figure 7 It is a schematic diagram of the software solving and extracting straight line features.

[0081] Figure 8 This is a single power line matching diagram.

[0082] Fig. 9 It is a schematic diagram of the three-dimensional coordinates of the point cloud obtained by separating the matched power line point cloud data. DETAILED DESCRIPTION

[0083] Below, in conjunction with the accompanying drawings, the technical solution of the transmission line fitting method for LiDAR robot point cloud power inspection provided by the present application is further described so that technicians in this field can better understand the present application and implement it.

[0084] This application introduces the LiDAR system and configures it on the line patrol robot, corrects the positioning, point cloud measurement accuracy and error analysis of the LiDAR system, and uses point cloud data to extract the power line equation;

[0085] 1. Combined with the high-voltage transmission line patrol robot platform, the working principle of the airborne lidar measurement system is explained, and a complete workflow for using the airborne lidar measurement system for data collection and generating a point cloud model of the transmission line is formed.

[0086] 2. The error sources of airborne lidar measurement are studied, and a variety of error models are constructed in detail. Combined with the characteristics of high-voltage transmission line patrol robots, the influence of various errors on the final laser foot point coordinate accuracy is quantitatively analyzed. It is found that the comprehensive error will significantly reduce the accuracy of the laser foot point coordinates, but the impact on it is not a simple linear superposition of various errors. The influence mechanism is very complex, and it is necessary to integrate multiple error sources before the system can be integrated and calibrated.

[0087] 3. In view of the inspection characteristics of high-voltage transmission line patrol robots, the airborne lidar measurement equipment is calibrated in flight. By correcting the roll angle and heading angle, the point cloud model image generated by the measurement is finally made clear, thereby improving the relative accuracy.

[0088] 4. Due to the limitations of the overhead line catenary equation in practical applications, a fitting method for the overhead line catenary equation based on the LiDAR robot system is proposed. Combined with the point cloud model of the transmission line generated by the airborne laser radar measurement equipment, the elevation projection method is used for filtering, the Canny operator edge detection and the improved Hough line detection method are used to extract the power lines. Finally, the point cloud data of the power lines are used to fit the power line equation, and it is concluded that it is more accurate and faster to generate the power line equation using the point cloud model.

[0089] 1. Power line equation fitting framework

[0090] The sag and line length of overhead power lines are key parameters, and traditional methods cannot accurately measure these parameters. The catenary equation of overhead power lines is used to calculate the shape of power lines. The integral form of the catenary equation of overhead power lines is:

[0091]

[0092] where σ 0 is the horizontal stress of the overhead line, γ is the specific load, C 1 , C 2 It is a constant whose value is related to the origin of the coordinate system.

[0093] The uneven terrain or the different heights of the towers will lead to unequal heights of the suspension points. The equation of the overhead line with unequal height suspension points is used as the mathematical model of the overhead ground line. The height difference between the two suspension points A and B is the height difference h, and the angle with the horizontal plane is the height difference angle Φ. Point A is taken as the origin of the coordinate system, such as Figure 1 :

[0094]

[0095] in, It is expressed as the length of the catenary in the overhead line with equal height suspension points;

[0096] The catenary shape of the overhead line As long as the ratio is the same, the final suspension curve of the overhead line is the same, and as long as The final suspension curve of the overhead line is certain, where y is the only influencing factor.

[0097] The horizontal stress is determined based on the ultimate strength of the conductor, taking into account the increase in stress at the hanging point, acceptable extreme weather conditions, and safety factors. The horizontal stress will be affected by the final elastic coefficient, temperature expansion coefficient, cross-sectional area, and temperature of the overhead line. At the same time, the specific load γ changes in different climatic environments, which ultimately causes the overhead line to present different sags, that is, the position of each point on the overhead line changes. The power line equation obtained using the overhead line catenary equation is an equation under specific conditions and has errors compared to the actual situation.

[0098] The power lines are scanned using a laser radar measurement system to generate a three-dimensional point cloud model of the power lines. The point cloud model is used to accurately determine the spatial three-dimensional coordinates of any point on the power lines and the specific shape of the power lines. A power line equation consistent with the actual situation is generated and fed back to the overhead line catenary equation, thereby providing accurate data parameters for power line research.

[0099] 2. Feature extraction of LiDAR power line scanning data

[0100] (1) Elevation distribution feature extraction

[0101] Based on the power line structure ( Figure 2 ), there is a clear height between the power lines and the ground, to ensure the safety of people and animals walking under the high-voltage overhead transmission lines, and to prevent them from being harmed by the high-voltage overhead transmission lines. That is, within a certain range, it is necessary to ensure that there is a difference between the power lines and the ground elevation, so the elevation difference can be used to make a preliminary distinction between ground points and power lines (including transmission line towers).

[0102] from Figure 3 It can be seen that among the measured data points, most of the data points are concentrated in the area with smaller elevation values, which are ground points. Only a small number of data points are concentrated in the area with larger elevation values, which are power line points. The points scattered between the ground points and the power line points are tower points and other ground points. Figure 2 and Figure 3 It can also be studied and concluded that within a certain interval, the power lines have basically the same elevation distribution or are similarly on the same horizontal plane, which is the basis of the algorithm of this application.

[0103] (2) Plane projection feature extraction

[0104] The point cloud data obtained after filtering include the following: power line point cloud data, pole tower point cloud data, and some unfiltered ground object point cloud data. Among these point cloud data, the unfiltered ground objects may be close to the power lines, which is less than the safe distance and easily leads to danger. The projection of the point cloud data on the horizontal plane after filtering shows that: the power lines appear as parallel lines spliced ​​together by discrete discontinuous points, which are extracted by linear fitting to obtain power line extraction data; pole towers and possible dangerous ground objects are irregular closed figures, which are separated by mathematical morphology methods. In addition, possible dangerous ground objects are vegetation or buildings, and the projected area is smaller than that of pole towers. Therefore, an area critical value is set to separate the two and distinguish between pole towers and possible dangerous ground objects.

[0105] 3. Power line fitting algorithm based on elevation projection

[0106] Initially, the filtering analysis was completed using the elevation critical value segmentation method. First, some ground points were removed, and the remaining data obtained were power line point cloud data, pole tower point cloud data, and some ground object point cloud data that were not filtered out. Then, the above data were converted into two-dimensional elevation value projections through elevation projection and resampling methods, and the power lines were extracted and fitted using a linear fitting method to obtain power line related data. Finally, the obtained image was processed to separate pole towers and possible dangerous objects.

[0107] (1) Filtering

[0108] According to the distribution characteristics of the data, the power line point cloud data is obtained, and then the elevation critical value segmentation method is used to remove the ground points, reduce the data and improve the fitting degree and fitting efficiency. The optimal iterative critical value method is used to achieve filtering. The specific calculation steps are as follows:

[0109] Step 1: Select the average elevation T 0 As the initial critical value of this calculation, that is, T k =T 0 ;

[0110] Step 2: According to the critical value T k Then the data is divided into two subsets, and the average elevation of the two subsets is calculated in turn, which are T A and T B , assuming T A <T k , T B >T k ;

[0111] Step 3: Get the critical value T according to formula 3 k+1 :

[0112]

[0113] Step 4: T k+1 Instead of T k ;

[0114] Step 5: Repeat the above process until T k Convergence, numerically T k+1 =T k , then T k is the selected segmentation threshold T;

[0115] Step 6: Divide the data into two parts according to the final segmentation threshold T obtained;

[0116] Iteratively filter out ground points. Figure 4 These are the ground points that are filtered out after analysis. Among these points, the gray points represent non-ground points, and the black points represent ground points. Some electric tower points in the figure are often filtered out as ground points, but they will not affect the subsequent analysis results.

[0117] (2) Elevation projection and resampling

[0118] The elevation value image is obtained by using point cloud projection and resampling methods, and the coverage and elevation range of the dataset on the XOY plane are calculated; the non-ground points are mapped to the XOY plane, and the projection sampling distance d is calculated according to the width W of the elevation value image by formula 4:

[0119]

[0120] The plane coordinates (x, y) of a point are converted into the corresponding grid row and column numbers (r, c) by equation 5:

[0121]

[0122] Resampling is applied point by point. If there is no laser point in the grid point, the gray value is zero. If there is a laser point in the grid point, the maximum elevation value is normalized to the gray range of [a, b] using the formula and regarded as the gray value gray:

[0123]

[0124] Finally, the elevation value image is obtained, which realizes the transformation of point cloud data from three-dimensional space to two-dimensional plane.

[0125] (3) Power line extraction based on straight line detection

[0126] A point in the original image coordinate system corresponds to a straight line in the polar coordinate system. All points displayed as straight lines in the original coordinate system have the same slope and intercept, so these straight lines correspond to the same point in the polar coordinate system. After projecting each point in the original coordinate system to the polar coordinate system, see if there are any clustered points in the polar coordinate system. Such clustered points correspond to the straight lines in the original coordinate system. In the elevation value image of the LiDAR power line point cloud data, the power lines are similar to line segments composed of a large number of discrete points, rather than complete straight lines. Hough transform is used to detect LiDAR data to prevent the power line data from being too discrete.

[0127] However, the traditional Hough transform first decomposes the variable plane into many squares according to a certain step size, and then obtains the variable values ​​of multiple pixel points on the same straight line in the above plane according to the many-to-one mapping. If the operation effect falls within a square in the variable plane, the squares are accumulated to get 1, and the variable of the straight line in the image space is obtained. The exhaustive search method is not only large in computation and memory usage; and the square with the most votes in the variable domain is surrounded by squares with more votes, which can easily interfere with the detection process and cause false detection or missed detection of the straight line. Therefore, when using the straight line detection method to detect power lines, a more optimized detection algorithm should be used.

[0128] (4) Power line extraction based on improved Hough line detection algorithm

[0129] First, the original image is preprocessed to remove all noise points; then the object and the background are separated, and the object is refined based on this; then, the adjacent pixels in the refined image are clustered and analyzed, and then perceptual grouping is used to further subdivide the adjacent pixels so that they can be closer to a straight line; finally, the Hough transform is used to perform straight line detection on each segment, and the final fitting is used to obtain the detection effect.

[0130] ① Clustering of adjacent pixels

[0131] Normally, the straight line obtained is not a smooth straight line. This is because external noise will interfere with the smoothness of the straight line. At the same time, problems such as the performance of the image acquisition equipment will also interfere with the smoothness of the straight line, resulting in the image being not as smooth as expected. After filtering, segmenting, thinning, edge processing, etc., the resulting image is still an unsmooth straight line, a broken line composed of unsmooth straight lines, and short lines formed by noise. Therefore, if the traditional Hough transform is used to process the data, it will be affected by these line segments, short lines, etc., resulting in a decrease in detection accuracy. The clustering of adjacent pixel points in the image is to reduce the interference of short lines formed by noise on the detection. The specific steps are as follows:

[0132] Step 1: First select a point P in the binary image 0 , then determine whether the point is 1, that is, P 0 Is it a feature pixel? If the obtained P 0 ≠1, then select a point in the binary image again. If P 0 =1, then enter the next stage;

[0133] Step 2: Record the position of the obtained feature pixel point as P, put it into one end of the feature point set S+P→S, and set the pixel value of the point to P=0;

[0134] Step 3: Judge all the neighborhoods of point P to see if they are feature points. If the final judgment result is a feature point, repeat the second step. If the final judgment result is not a feature point, go to the next step.

[0135] Step 4: Re-align P 0 Search the data in the neighborhood of the point to determine whether these data have connected feature pixels. If the judgment result is yes, execute the fifth step. If the judgment result is no, terminate the execution.

[0136] Step 5: Put the feature points into the other end of the set P+S→S, and the rest of the operations are the same as step 2;

[0137] Step 6: Continue to determine whether these data are feature points. If the determination result is yes, execute step 5. If the determination result is no, terminate the execution.

[0138] Figure 5 It is the task graph of the clustering process of adjacent pixels. After the above processing, the pixels are aggregated. In addition, those independent pixels are removed and the pixels are regarded as similar to straight lines. This will further determine the detection object and separate the parallel lines with close intervals, preventing false detection and missed detection, and reducing the scale of computational data.

[0139] ② Perceptual grouping of straight lines

[0140] In order to facilitate the subsequent processing, the polyline is further subdivided, and the resulting polyline will be closer to a straight line, such as Figure 6 As shown (x 1 ,y 1 )、(x 2 ,y 2 ) represents the coordinates of the two end points of the short line, and B is the coordinates of the short line passing through (x 1 ,y 1 )、(x 2 ,y 2 ) of the straight line, (x 3 ,y3 ) represents the coordinates of the point where line segment A deviates from line B by a distance d, and d is calculated by formula 7:

[0141] d=[(y 3 -y 1 )(x 2 -x 1 )-(x 3 -x 1 )(y 2 -y 1 )] / L Formula 7

[0142] Where: make:

[0143] R=L / maxd Formula 8

[0144] The ratio R is only related to the line itself, and has nothing to do with the size of the image obtained, and does not change with the size of the image. The point with the largest offset distance is used for binary segmentation to obtain the ratio R i (i=1, 2), compare it with the ratio of the original line segment. If the ratio R i If the value of is larger, the original line segment is replaced by two sub-line segments; otherwise, the original line segment is retained. In this way, the original line segment is closer to a straight line and no subsequent processing is required.

[0145] Through calculation and analysis, each line segment will be closer to a straight line, determining the detected straight line and improving the detection accuracy.

[0146] ③Straight line segment detection based on free probability transformation

[0147] The approximate straight line segments are tested one by one, and two points (x p ,y p )、(x q ,y q ), use two points to generate a straight line, and the generated straight line variable is obtained by the following formula:

[0148]

[0149] ρ=x p cosθ+y p sinθ or ρ = x q cosθ+y q sinθ Equation 10

[0150] The use of clustering and perceptual grouping can better ensure effectiveness, so when performing Hough transform on the above image, it only needs to reach a very small critical value. The above analytical calculation results are as follows:

[0151] Step 1: Get all the pixels d i (x i ,y i ) is regarded as a set D;

[0152] Step 2: Select d from the above set D. p , d q Calculate whether they meet the requirements of a point pair. If the two selected d p , d q If the point does not meet the point pair requirement, reselect d p , d q ;

[0153] Step 3: Find d p , d q The corresponding point (θ, ρ) is accumulated to 1 at the corresponding position;

[0154] Step 4: If the value of a certain point in the parameter space is increased to the critical value, the above calculation is stopped, and the straight line corresponding to this point is the desired straight line.

[0155] Step 5: Use opencv software to process the elevation projection. The core code is:

[0156] if len(sys.argv)<3:

[0157] print("Usage: python hough-lines.py[img_path][output_path]")

[0158] sys.exit()

[0159] imgIn, imgOut=sys.argv[1].sys.argv[2]

[0160] img = cv2.imread(imgIn)

[0161] #img=cv2.imread('railroad.jpg')

[0162] gray=cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)

[0163] edges=cv2.Canny(gray, 100, 200)

[0164] lines=cv2.houghLines(edges, 1, np.pi / 180, 200)

[0165] Through software calculation, the projections of most power lines are detected. Figure 7 )

[0166] Sort the detected straight lines by their lengths p in the polar coordinate system, and manually select the most suitable power line matching line by comparing the straight lines corresponding to different p values. Figure 8 )

[0167] Finally, the matched power line point cloud data is separated to obtain the point cloud three-dimensional coordinates (such as Fig. 9 ), input the point cloud coordinates into MATLAB software to display its spatial position and form;

[0168] Finally, transform the X-axis, Y-axis, and Z-axis data, so that Y * =Z, The data is fitted and finally the equation of the point cloud data in the plane perpendicular to the power lines is obtained.

Claims

1. A transmission line fitting method for LiDAR robot point cloud power inspection, characterized in that: The LiDAR system is introduced and configured on the line patrol robot. Based on the positioning of the LiDAR system, point cloud measurement accuracy and error analysis correction by the intelligent robot, the power line equation is extracted using point cloud data. S1: Build a LiDAR robot power inspection platform mounted on an intelligent robot, and form a complete workflow for data collection and generation of a point cloud model of a power transmission line using an airborne LiDAR measurement system; S2: Establish multiple error models for configuring the LiDAR system on the line patrol robot, quantitatively analyze the impact of multiple errors on the final laser foot point coordinate accuracy, significantly reduce the laser foot point coordinate accuracy based on the comprehensive error, and integrate multiple error sources to perform integrated calibration of the system; S3: Based on the inspection features of the intelligent robot for power transmission lines, the airborne laser radar measurement equipment is calibrated in flight. By correcting the roll angle and heading angle, the point cloud model image generated by the measurement is made clear, thereby improving the relative accuracy. S4: Establish a method for fitting the catenary equation of overhead lines based on the LiDAR robot system. Combined with the point cloud model of the transmission line generated by the airborne laser radar measurement equipment, a power line equation fitting framework is constructed. The elevation distribution features and plane projection features of the LiDAR power line scanning data are extracted. The elevation projection method is used for filtering, elevation projection and resampling, Canny operator edge detection and improved Hough line detection methods are used to extract power lines. After clustering of adjacent pixels, perception grouping of lines, and line segment detection based on free probability transformation, the power line equation is fitted using the point cloud data of the power line. It is concluded that generating the power line equation using the point cloud model is more accurate and faster.

2. The transmission line fitting method for LiDAR robot point cloud power inspection according to claim 1 is characterized in that: Power line equation fitting framework: The shape of the power line is calculated using the catenary equation of the overhead power line. The integral form of the catenary equation of the overhead power line is: Where σ0 is the horizontal stress of the overhead line, γ is the specific load, C1 and C2 are constants, and their values ​​are related to the origin of the coordinate system; The equation of overhead wire with unequal height suspension points is used as the mathematical model of overhead ground wire. The height difference between the two suspension points A and B is the height difference h, and the angle between the two suspension points and the horizontal plane is the height difference angle Φ. Point A is taken as the origin of the coordinate system: in, It is expressed as the length of the catenary in the overhead line with equal height suspension points; The catenary shape of the overhead line As long as the ratio is the same, the final suspension curve of the overhead line is the same, and as long as The final suspension curve of the overhead line is certain, where y is the only influencing factor; The power lines are scanned using a laser radar measurement system to generate a three-dimensional point cloud model of the power lines. The point cloud model is used to accurately determine the spatial three-dimensional coordinates of any point on the power lines and the specific shape of the power lines. A power line equation consistent with the actual situation is generated and fed back to the overhead line catenary equation, thereby providing accurate data parameters for power line research.

3. The transmission line fitting method for LiDAR robot point cloud power inspection according to claim 1 is characterized in that: Elevation distribution feature extraction: Based on the power line structure, within a certain range, it is ensured that there is a difference between the power line and the ground elevation, and the elevation difference is used to make a preliminary distinction between the ground points and the power line; Among the measured data points, the vast majority are concentrated in areas with smaller elevation values, which are ground points. Only a small number of data points are concentrated in areas with larger elevation values, which are power line points. Scattered between ground points and power line points are tower points and other ground features. Within a certain interval, power lines have basically the same elevation distribution or are similarly on the same horizontal plane.

4. The transmission line fitting method for LiDAR robot point cloud power inspection according to claim 1 is characterized in that: Plane projection feature extraction: The point cloud data obtained after filtering includes the following: power line point cloud data, pole tower point cloud data, and some unfiltered ground object point cloud data. Among these point cloud data, the unfiltered ground objects may be close to the power lines, which is less than the safe distance. The projection of the point cloud data on the horizontal plane after filtering shows that: power lines appear as parallel lines spliced ​​together by discrete discontinuous points, which are extracted by linear fitting to obtain power line extraction data; pole towers and possible dangerous ground objects are both irregular closed figures, which are separated by mathematical morphology. In addition, possible dangerous ground objects are vegetation or buildings, and their projected area is smaller than that of pole towers. An area critical value is set to separate the two and distinguish between pole towers and possible dangerous ground objects.

5. The transmission line fitting method for LiDAR robot point cloud power inspection according to claim 1 is characterized in that: Power line fitting algorithm based on elevation projection: Initially, the filtering analysis is completed by using the elevation critical value segmentation method. First, some ground points are removed, and the remaining data are power line point cloud data, tower point cloud data, and some ground point cloud data that are not filtered out; then, the above data are converted into two-dimensional elevation value projection through elevation projection method and resampling method, and the power lines are extracted and fitted by linear fitting method to obtain power line related data; finally, the obtained image is processed to facilitate the separation of towers and possible dangerous ground objects.

6. The transmission line fitting method for LiDAR robot point cloud power inspection according to claim 1, characterized in that: Filtering of the power line fitting algorithm of elevation projection: According to the distribution characteristics of the data, the power line point cloud data is obtained, and then the elevation critical value segmentation method is used to remove the ground points, reduce the data and improve the fitting degree and fitting efficiency. The optimal iterative critical value method is used to achieve filtering. The specific calculation steps are as follows: Step 1: Select the average elevation T0 as the initial critical value for this calculation, that is, T k =T0; Step 2: According to the critical value T k , and then divide the data into two subsets, and calculate the average elevation of the two subsets, which are T and A and T B , assuming T A <T k , T B >T k ; Step 3: Get the critical value T according to formula 3 k+1 : Step 4: T k+1 Instead of T k ; Step 5: Repeat the above process until T k Convergence, numerically T k+1 =T k , then T k is the selected segmentation threshold T; Step 6: Based on the final segmentation threshold T obtained, the data is divided into two parts.

7. The transmission line fitting method for LiDAR robot point cloud power inspection according to claim 1 is characterized in that: Elevation projection and resampling: The point cloud projection and resampling method is used to obtain the elevation value image, and the coverage and elevation range of the dataset on the XOY plane are calculated; the non-ground points are mapped to the XOY plane, and the projection sampling distance d is calculated according to the width W of the elevation value image through formula 4: The plane coordinates (x, y) of a point are converted into the corresponding grid row and column numbers (r, c) by equation 5: Resampling is applied point by point. If there is no laser point in the grid point, the gray value is zero. If there is a laser point in the grid point, the maximum elevation value is normalized to the gray range of [a, b] using the formula and regarded as the gray value gray: Finally, the elevation value image is obtained, which realizes the transformation of point cloud data from three-dimensional space to two-dimensional plane.

8. The transmission line fitting method for LiDAR robot point cloud power inspection according to claim 1, characterized in that: Clustering of adjacent pixels: After filtering, segmenting, thinning, edge processing, etc. of the actual image, cluster the adjacent pixels in the image to reduce the interference of short lines formed by noise on the detection. The specific steps are as follows: Step 1: First, select a point P0 in the binary image, and then determine whether the point is 1, that is, whether P0 is a feature pixel. If the obtained P0≠1, select a point in the binary image again. If P0=1, enter the next stage; Step 2: Record the position of the obtained feature pixel point as P, put it into one end of the feature point set S+P→S, and set the pixel value of the point to P=0; Step 3: Judge all the neighborhoods of point P to see if they are feature points. If the final judgment result is a feature point, repeat the second step. If the final judgment result is not a feature point, go to the next step. Step 4: Re-search the data in the neighborhood of point P0 to determine whether these data have connected feature pixels. If the result is yes, execute step 5. If not, terminate the execution. Step 5: Put the feature points into the other end of the set P+S→S, and the rest of the operations are the same as step 2; Step 6: Continue to determine whether these data are feature points. If the determination result is yes, execute step 5. If the determination result is no, terminate the execution. After the above processing, the pixels are aggregated. In addition, the independent pixels are removed and the pixels are regarded as similar to straight lines. This will further determine the detection object and separate the parallel lines with close intervals.

9. The transmission line fitting method for LiDAR robot point cloud power inspection according to claim 1, characterized in that: Perceptual grouping of straight lines: Further subdividing the polyline will make it closer to a straight line. (x1, y1) and (x2, y2) represent the coordinates of the two end points of the short line, and B is the straight line passing through (x1, y1) and (x2, y2). (x3, y3) represents the coordinates of the point where the line segment A deviates from the straight line B by a distance d. d is calculated by formula 7: d=[(y3-y1)(x2-x1)-(x3-x1)(y2-y1)] / L Formula 7 Where: make: R=L / maxd Formula 8 The ratio R is only related to the line itself, and has nothing to do with the size of the image obtained, and does not change with the size of the image. The point with the largest offset distance is used for binary segmentation to obtain the ratio R i (i=1, 2), compare it with the ratio of the original line segment. If the ratio R i If the value of is larger, the original line segment is replaced by two sub-line segments; Otherwise, the original line segment is retained, so that the original line segment is closer to a straight line and no further processing is required; Through calculation and analysis, each line segment will be closer to a straight line, determining the detected straight line and improving the detection accuracy.

10. The transmission line fitting method for LiDAR robot point cloud power inspection according to claim 1, characterized in that: Straight line segment detection based on free probability transformation: detect the approximate straight line segments in turn, and select any two points (x p ,y p )、(x q ,y q ), use two points to generate a straight line, and the generated straight line variable is obtained by the following formula: ρ = x p cosθ + y p sinθ or ρ = x q cosθ + y q sinθ Equation 10 The use of clustering and perceptual grouping can better ensure effectiveness, so when performing Hough transform on the above image, it only needs to reach a very small critical value. The above analytical calculation results are as follows: Step 1: Get all the pixels d i (x i ,y i ) is regarded as a set D; Step 2: Select d from the above set D. p , d q Calculate whether they meet the requirements of a point pair. If the two selected d p , d q If the point does not meet the point pair requirement, reselect d p , d q ; Step 3: Find d p , d q The corresponding point (θ, ρ) is accumulated to 1 at the corresponding position; Step 4: If the value of a certain point in the parameter space reaches the critical value, the above calculation is stopped, and the straight line corresponding to this point is the desired straight line; Step 5: Use OpenCV software to process the elevation projection. Through software calculation, the projections of most power lines are detected. Sort the detected straight lines by their lengths p in the polar coordinate system, and manually select the power line matching line that best meets the needs by comparing the straight lines corresponding to different p values; Finally, the matching power line point cloud data is separated to obtain the three-dimensional coordinates of the point cloud, and the point cloud coordinates are input into the MATLAB software to display its spatial position and form.

Images