Tower feature matching method and system

Abstract

The invention relates to the technical field of power system safety detection, in particular to a tower feature matching method and system, and the method comprises the steps: carrying out the preprocessing of collected tower point cloud data, extracting target point cloud data, and enabling the target point cloud data to be point cloud data with a tower as the center; tower structure analysis and feature detection are carried out on the target point cloud data, and tower features are extracted; classifying tower features, and dividing normal samples and abnormal samples; and according to the abnormal sample, a path optimization algorithm is adopted to dynamically plan an operation path of an operator. The method has the beneficial effects that the automation level of tower detection is improved, manual intervention is reduced, the detection accuracy and the operation safety are improved, and a powerful guarantee is provided for stable operation of a power system.

Description

Technical Field

[0001] The present invention relates to the technical field of power system safety detection, and particularly to a tower feature matching method and system. Background Art

[0002] With the rapid development of the power transmission network and the continuous expansion of the infrastructure construction scale, transmission towers, as an indispensable and important part of the power system, the stability and safety of their operating states are crucial for the normal operation of the power grid. Transmission towers are long-term exposed to complex natural environments and are vulnerable to various factors such as wind, rain, snow, corrosion, and mechanical fatigue, resulting in structural deformation, crack generation, or corrosion damage. To ensure the stability of power equipment, traditional tower detection and maintenance mainly rely on manual inspections. Technicians judge the abnormal conditions of towers through visual inspections, measuring tools, or simple equipment. However, manual inspections have limitations such as high labor intensity, low efficiency, and insufficient detection accuracy, and are difficult to meet the requirements of intelligent and refined management of modern power grids.

[0003] In recent years, with the progress of technologies such as three-dimensional laser scanning, machine learning, and unmanned aerial vehicle mapping, tower state detection based on point cloud data has gradually become a research hotspot. Point cloud data can accurately obtain the three-dimensional spatial information of towers and provide high-resolution geometric shapes and structural features. Existing point cloud-based tower detection methods still have many deficiencies: there is a lot of noise in point cloud data, and it is difficult to guarantee the accuracy of data preprocessing and alignment, which affects the accuracy of subsequent analysis; the tower segmentation and feature extraction methods are not yet perfect, and it is difficult to comprehensively identify the inclination angle, cracks, and corrosion areas of towers; there is a lack of effective anomaly detection models, and existing technologies often rely on a single algorithm for predicting tower structure deviations and are difficult to achieve high-precision and multi-dimensional deviation identification. In terms of operation safety guarantee, existing technologies rarely focus on dynamic path optimization based on real-time detection results, resulting in the operation efficiency and safety of operators in high-risk areas being difficult to be fully guaranteed. Summary of the Invention

[0004] To solve the above technical problems, the present invention provides the following technical solutions:

[0005] In a first aspect, the present invention provides a tower feature matching method, including preprocessing the collected tower point cloud data to extract target point cloud data, where the target point cloud data is the point cloud data centered on the tower;

[0006] Performing structural analysis and feature detection of the tower on the target point cloud data to extract tower features;

[0007] Classifying the tower features to divide normal samples and abnormal samples;

[0008] The operation route of the operator is dynamically planned using a path optimization algorithm based on abnormal samples.

[0009] As a preferred embodiment of the pole tower feature matching method of the present invention, wherein: the preprocessing of the collected pole tower point cloud data includes:

[0010] Using a drone equipped with a sensor to collect pole tower point cloud data;

[0011] Checking the integrity of the collected pole tower point cloud data, and removing error data and abnormal points;

[0012] Aligning the collected pole tower point cloud data and converting it to a unified coordinate system.

[0013] 4. As a preferred embodiment of the pole tower feature matching method of the present invention, wherein: the structural analysis and feature detection of the pole tower for the target point cloud data include:

[0014] Dividing the target point cloud data into blocks of tower feet, tower bodies, and tower tops according to the physical structure of the pole tower;

[0015] Calculating the geometric center and tilt angle of each block;

[0016] Detecting cracks and corrosion areas for each block.

[0017] As a preferred embodiment of the pole tower feature matching method of the present invention, wherein: classifying the pole tower features and dividing normal samples and abnormal samples include:

[0018] Normalizing the pole tower features according to the manually labeled sample tags;

[0019] Using a random forest model to classify the pole tower features and divide normal samples and abnormal samples.

[0020] As a preferred embodiment of the pole tower feature matching method of the present invention, wherein: dynamically planning the operation route of the operator using a path optimization algorithm based on abnormal samples includes:

[0021] Dividing the operation area into three-dimensional grid cells and defining risk areas;

[0022] Setting a cost function H for each three-dimensional grid cell, and the cost function H comprehensively considers the distance from the center of the three-dimensional grid cell to the risk area and the height cost;

[0023] Using the D* algorithm to dynamically optimize the operation route of the operator according to the cost function H.

[0024] As a preferred embodiment of the pole tower feature matching method of the present invention, wherein: the method further includes evaluating the confidence level of the division result.

[0025] As a preferred solution of the pole and tower feature matching method of the present invention, wherein: the pole and tower features include but are not limited to the tilt angle, the number and size of cracks, and the area of the corrosion region.

[0026] In a second aspect, the present invention provides a pole and tower feature matching method, including: a processing module for preprocessing the collected pole and tower point cloud data to extract target point cloud data, where the target point cloud data is the point cloud data centered on the pole and tower;

[0027] An analysis and detection module for performing structural analysis and feature detection of the pole and tower on the target point cloud data to extract pole and tower features;

[0028] A classification module for classifying the pole and tower features to divide normal samples and abnormal samples;

[0029] An optimization module for dynamically planning the operation route of the operator according to the abnormal samples by using a path optimization algorithm.

[0030] In a third aspect, the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, the steps of the above method are implemented.

[0031] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0032] Compared with the prior art, the beneficial effects of the present invention are as follows: the point cloud data of the pole and tower can be collected in real time by using the sensor carried by the unmanned aerial vehicle; through preprocessing and feature extraction, the accurate analysis of the structural state of the pole and tower can be realized, including the tilt angle, cracks and corrosion regions; the random forest model is used to classify the pole and tower features, which can effectively identify normal and abnormal samples, and the artificial review can be guided by quantifying the prediction confidence level, and for abnormal samples, the D* algorithm is used to dynamically optimize the route of the operator to avoid high-risk areas. The present invention not only improves the automation level of pole and tower detection, but also reduces manual intervention, improves the accuracy of detection and operation safety, and provides a strong guarantee for the stable operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0034] Figure 1 For the pole and tower feature matching method. Detailed implementation manners

[0035] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following provides a detailed description of the specific implementation manners of the present invention with reference to the accompanying drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0036] Embodiment 1. Refer to Figure 1 , which is the first embodiment of the present invention. This embodiment provides a method for matching tower characteristics, which includes:

[0037] S1. Preprocess the collected tower point cloud data to extract target point cloud data, where the target point cloud data is the point cloud data centered on the tower.

[0038] Further, the preprocessing of the collected tower point cloud data includes:

[0039] Use a drone to carry a sensor to collect tower point cloud data;

[0040] It should be noted that the carried sensor includes but is not limited to GPS (Global Positioning System), inertial measurement unit (IMU), and lidar (LiDAR). The collected tower point cloud data is specifically obtained by collecting the point cloud data of each reflection point around the tower in real time.

[0041] Check the integrity of the collected tower point cloud data, and remove error data and abnormal points;

[0042] It should be noted that abnormal point cloud data is detected and removed by using the Z-score method. For the edge area of the tower, the density calculation formula based on DBSCAN is used to define the point density as the number of points in the neighborhood within the unit spherical radius, and isolated points are detected and removed by detecting local density changes. For the data missing part caused by removing abnormal points or isolated points, linear interpolation is used to fill the removed missing values.

[0043] Align the collected tower point cloud data and convert it to a unified coordinate system.

[0044] It should be noted that through the GPS and inertial measurement unit in the drone, while collecting the tower point cloud data, the GPS data of the sampling points is recorded in real time, the GPS data is converted to a unified geographic coordinate system, a preliminary alignment model is generated using the GPS data and the geometric characteristics of the tower structure (such as the center line of the tower body), and the iterative closest point algorithm is used to optimize the alignment for the preliminary alignment model.

[0045] Further, it should be noted that by taking the center point of the pole tower as the reference, spatially cropping the point cloud data of the pole tower in the unified (global) coordinate system, retaining the target point cloud data, and further compressing the point cloud data by using voxel grid downsampling, and identifying the ground points by using the height distribution (Z-axis value) of the point cloud data, removing the background objects around the pole tower through density and curvature analysis, the extraction of the target point cloud data is realized.

[0046] Preferably, this step improves the accuracy and efficiency of point cloud data processing, providing high-quality data support for the subsequent extraction of pole tower features and structural analysis.

[0047] S2. Conduct structural analysis and feature detection of the pole tower on the target point cloud data, and extract the pole tower features.

[0048] Furthermore, the pole tower features include, but are not limited to, the inclination angle, the number and size of cracks, and the area of the corrosion region.

[0049] Furthermore, conducting structural analysis and feature detection of the pole tower on the target point cloud data includes:

[0050] Dividing the target point cloud data into blocks of tower feet, tower body, and tower top according to the physical structure of the pole tower;

[0051] It should be noted that manually divide the target point cloud data into blocks of tower feet, tower body, and tower top according to the height distribution. The tower feet refer to the part where the pole tower directly contacts the foundation, which is the support foundation of the pole tower and usually includes foundation bolts, the foundation structure buried underground, and the bottom tower frame connected to the ground; the tower body refers to the middle part of the pole tower, connecting the tower feet and the tower top, which is the main load-bearing structure of the pole tower. The tower body is usually composed of metal rods, support frames, or integral columns; the tower top is the uppermost part of the pole tower, usually equipped with various accessory devices (such as antennas, optical cables, lightning rods, etc.).

[0052] Calculate the geometric center and inclination angle of each block;

[0053] It should be noted that based on the ground points, redefine the coordinate system, take the ground height as the zero value of the z-axis, ensure that the z-values of all target point cloud data are positive, and convert the z-values of all target point cloud data. Finally, calculate the geometric center coordinates of all target point cloud data in each block. The specific formula is:

[0054]

[0055] Among them, G(x g , y g , z g ) represents the geometric center coordinate g in block G;

[0056] x i 、y i, z i represents the three-dimensional spatial coordinates of the i-th point cloud data in block G;

[0057] N represents the number of point cloud data within block G.

[0058] Connect the geometric center coordinates of all blocks to fit the center axis of the pole tower. Further divide the center axis of the pole tower into two sections from the tower foot to the tower body and from the tower body to the tower top, and calculate the horizontal offset from the tower foot to the tower body and from the tower body to the tower top respectively. The formula is:

[0059]

[0060] where, ΔL base-mid and ΔL mid-top represent the horizontal offsets from the tower foot to the tower body and from the tower body to the tower top respectively;

[0061] x base and y base represent the x-axis and y-axis coordinate values of the geometric center coordinates of the tower foot;

[0062] x mid and y mid represent the x-axis and y-axis coordinate values of the geometric center coordinates of the tower body;

[0063] x top and y top represent the x-axis and y-axis coordinate values of the geometric center coordinates of the tower top.

[0064] Calculate the height difference from the tower foot to the tower body and from the tower body to the tower top. The formula is:

[0065] h base-mid = |z mid - z base |

[0066] h mid-top = |z top-zmid |

[0067] where, Z base , Z mid , Z top represent the z-axis coordinate values of the geometric center coordinates of the tower foot, tower body, and tower top respectively.

[0068] According to the obtained horizontal offset and height difference, calculate the inclination angle of each section of the center axis of the pole tower. The inclination angle is a reasonable physical quantity obtained by converting the spatial coordinate offset and height difference. The formula is:

[0069]

[0070] where, θ base-mid represents the inclination angle from the tower foot to the tower body;

[0071] θ mid-top Represents the inclination angle from the tower body to the tower top.

[0072] Calculate the overall inclination angle of the pole tower according to the weighted average of the sectional inclination angles. The formula is:

[0073]

[0074] Where, θ total Represents the overall inclination angle of the pole tower;

[0075] w base-mid and w mid-top Represent the weights of the inclination angles from the tower foot to the tower body and from the tower body to the tower top respectively. The expression is:

[0076]

[0077] This weight combines the horizontal offset and the height difference, and can more comprehensively reflect the actual inclination contribution of each section. Paragraphs with large offsets but small height differences will not be ignored, and paragraphs with large heights but small offsets can also be reasonably reflected.

[0078] Preferably, in this step, by manually dividing blocks and redefining the coordinate system, the clear distribution and height consistency of the point cloud data of each part are ensured; through geometric center calculation, central axis fitting, and comprehensive analysis of horizontal offset and height difference, the spatial position and inclination state of each part of the pole tower are accurately reflected. By using the weighted average of sectional inclination angles, the overall inclination of the pole tower is comprehensively measured, overcoming the limitations of single-angle analysis, and providing an efficient and reliable calculation method for the structural monitoring and safety assessment of the pole tower.

[0079] Detect cracks and corrosion areas for each block.

[0080] It should be noted that the target point cloud data of the selected blocks of the tower body and the tower top are analyzed. The normal vectors of each target point cloud data are calculated using the point cloud normal vector estimation, and the change rate of the normal vectors of the target point cloud data and its nearest neighbor points is calculated. The formula is:

[0081]

[0082] Where, n i，x , n i，y , n i，z Represent the values of the normal vector of the i-th target point cloud data in the block on the x-axis, y-axis, and z-axis respectively;

[0083] n j，x , n j，y , n j，zThey respectively represent the values of the normal vector of the j-th target point cloud data in the block on the x-axis, y-axis, and z-axis.

[0084] According to the obtained normal vector change rate, a change rate threshold is set. If the normal vector change rate between the target point cloud data and its nearest neighbor point is greater than the change rate threshold, the target point cloud data is marked as a crack candidate point; otherwise, no operation is performed. All crack candidate points are combined to form a crack candidate area and cropped into a depth map, and the depth map is input into the YOLOv8 deep learning model for crack detection and precise annotation to obtain the depth map bounding box of each crack area. Among them, the YOLOv8 deep learning model is an object detection and segmentation algorithm. YOLOv8 adopts a unified framework and can simultaneously handle object detection, instance segmentation, and semantic segmentation tasks. Users only need to adjust the parameters through the configuration file. Further, according to the pixel length of the depth map bounding box and combined with the resolution of the target point cloud data, the actual length and width of the crack are calculated.

[0085] Furthermore, it should be noted that the surface of the pole tower usually has a regular structure along the z direction, such as a smooth cylindrical surface and a rectangular columnar surface. If there is corrosion or indentation, its characteristic is often a local height reduction (smaller z value), rather than a significant change in the horizontal position. Therefore, the z value of the point cloud data in the corrosion area of the pole tower will be lower than the average value of the z values of its neighboring points. By extracting the point cloud data on the surface of the pole tower and calculating the local surface depression depth, the formula is:

[0086] D c =|z δ -z c |

[0087] Among them, D c represents the local surface depression depth of the c-th target point cloud data on the surface of the pole tower;

[0088] z δ represents the average value of the z values of the local surface target point cloud data on the surface of the pole tower;

[0089] z c represents the z value of the c-th target point cloud data on the surface of the pole tower.

[0090] Record the maximum local surface depression depth in the target point cloud data. According to the obtained local surface depression depth, set a depth threshold. If the local surface depression depth of the target point cloud data is greater than the depth threshold, mark the target point cloud data as a corrosion point, cluster the corrosion points into a corrosion area, use the DBSCAN algorithm to identify the corrosion area and calculate the area of the corrosion area. The DBSCAN algorithm refers to a density-based clustering algorithm that can discover clusters of any shape and has strong robustness to noise data. The main idea of the DBSCAN algorithm is to divide points into core points, boundary points, and noise points by analyzing the density distribution of the data points, thus forming clusters.

[0091] Finally, statistically analyze the overall tilt angle, number of cracks, average crack length, average crack width, number of corrosion areas, average corrosion area, and maximum local surface depression depth of the pole tower and use them as pole tower features.

[0092] Preferably, in this step, through point cloud normal vector estimation and rate of change analysis, the crack candidate points of the pole tower can be accurately identified, and combined with the YOLOv8 deep learning model, the crack area can be detected and accurately labeled to obtain the actual length and width of the cracks. At the same time, by calculating the local surface depression depth and combining with the DBSCAN clustering algorithm, the corrosion area and area distribution of the pole tower can be accurately identified, effectively solving the problems of low efficiency and insufficient accuracy in crack and corrosion identification in traditional detection methods, realizing the efficient extraction of key features of the pole tower (such as tilt angle, crack, and corrosion features), providing high-quality input data for pole tower condition monitoring and subsequent deviation analysis, and significantly improving the intelligence and accuracy of detection.

[0093] S3. Classify the pole tower features and divide them into normal samples and abnormal samples.

[0094] Furthermore, classifying the pole tower features and dividing them into normal samples and abnormal samples includes:

[0095] Normalize the pole tower features according to the manually labeled sample labels. The sample labels are obtained by manually labeling the pole tower features based on the pole tower design drawings. Normalize all the pole tower features to facilitate the training of the model, and divide the normalized pole tower features and sample labels into a training set and a test set.

[0096] It should be noted that the manually standard sample labels refer to the technical personnel manually labeling normal samples and abnormal samples for all pole tower features. The normal sample label is 0, and the abnormal sample label is 1.

[0097] Use a random forest model to classify the pole tower features and divide them into normal samples and abnormal samples.

[0098] It should be noted that a random forest model is used as the deviation prediction model and decision trees are constructed (constructing decision trees means setting the number of decision trees. For each decision tree, a subset of the training set is randomly selected), the feature selection algorithm (such as the Gini index) is used to split the data to generate decision tree nodes, the maximum split threshold is set, and if the tree depth of the decision tree reaches the maximum split threshold, the split stops. The validation set is used to evaluate the random forest model, and grid search is used to adjust the hyperparameters to optimize the performance of the random forest model. Specifically, the new tower pole features are input into the trained random forest model, the deviation probability is calculated according to the voting ratio of each decision tree, and the sample label and abnormal classification probability μ of the tower pole are output f 。

[0099] Furthermore, it should be noted that the random forest model is used to score the importance of each tower pole feature, and the top M tower pole features with high importance are selected as the reasons for the tower pole deviation. The value of M is determined according to the actual situation; classification processing is performed based on the sample label. If it is determined that the tower pole feature is a normal sample, it is marked that there is no obvious abnormality in the tower pole structure. If it is determined that the tower pole feature is an abnormal sample, the alarm mechanism is triggered, the operation route is dynamically optimized for the operator, and the reasons for the tower pole deviation in the classification processing operation steps are sent to the technical personnel

[0100] Furthermore, the method also includes evaluating the confidence level of the division result

[0101] It should be noted that for the abnormal classification probability μ f quantitative prediction confidence is performed as follows

[0102] Set a low confidence threshold μ1 and a high confidence threshold μ2, and μ2 > μ1, to guide manual review, specifically including

[0103] If μ f > μ2, it means that the random forest model determines that there is a deviation in the tower pole feature, and a review is required to ensure that high-risk problems are resolved in a timely manner and major errors are prevented from being missed

[0104] If μ1 < μ f ≤ μ2, it means that the confidence level of the tower pole feature is not high, and further manual review is required and the review priority is reduced to save time and resources

[0105] If μ f ≤ μ1, it means that the random forest model determines that the tower pole feature is normal and has a high confidence level, and no review is required and classification processing operations are performed to reduce unnecessary manual review workload

[0106] Preferably, in this step, a random forest model is used to predict the deviation of the tower characteristics and optimize the classification review. Combined with the manually labeled normal and abnormal samples for training, a high-performance decision tree is constructed using a feature selection algorithm, realizing the accurate calculation of the deviation probability and the confidence quantification evaluation of the classification results. Through the direct marking of high-confidence samples and the priority review of low-confidence samples, the efficiency and accuracy of anomaly detection are significantly improved, major risk omissions are avoided, and unnecessary manual review workload is reduced. Based on the feature importance score, the deviation reason is clarified, and the alarm mechanism and dynamic path optimization are triggered, effectively improving the intelligent level of tower maintenance and providing accurate decision-making support for technicians.

[0107] S4. Dynamically plan the operation route of the operator according to the abnormal samples using a path optimization algorithm.

[0108] Further, dynamically planning the operation route of the operator according to the abnormal samples using a path optimization algorithm includes:

[0109] Divide the operation area into three-dimensional grid cells and define risk areas. Among them, each three-dimensional grid cell represents a cubic area, and the size of the three-dimensional grid cell is determined according to the operation requirements;

[0110] It should be noted that the crack area and the corrosion area are defined as risk areas.

[0111] Set a cost function H for each three-dimensional grid cell. The cost function H comprehensively considers the distance from the center of the three-dimensional grid cell to the risk area and the height cost. Among them, when the three-dimensional grid cell is within the risk area, mark the three-dimensional grid cell as an impassable state;

[0112] It should be noted that the cost function H has the formula:

[0113]

[0114] Among them, ω1, ω2, and ω3 respectively represent weight coefficients, and the weight coefficients are initially set according to the actual scenario requirements;

[0115] E r represents the distance from the center of the three-dimensional grid cell to the crack area;

[0116] E k represents the distance from the center of the three-dimensional grid cell to the corrosion area;

[0117] b represents the height cost, which is obtained by calculating the height coordinates of two adjacent three-dimensional grid cells in the path.

[0118] It should be further noted that if the three-dimensional grid cells are too close to the crack or corrosion area, the cost will increase rapidly, so that the path planning tries to avoid these high-risk areas as much as possible. In the height cost in path planning, it refers to the difficulty of the operator's movement in the vertical direction, which reflects the extra consumption required by the operator in the case of a large height difference. The introduction of the height cost is to be closer to the actual operation conditions, avoid too many vertical movements in path planning, and thus improve the operability and efficiency of the path.

[0119] The D* algorithm is used to dynamically optimize the operation route of the operator according to the cost function H.

[0120] It should be noted that the cumulative cost of the operator from the starting three-dimensional grid cell to the current three-dimensional grid cell is calculated, and the Euclidean distance from the center of the current three-dimensional grid cell to the center of the target three-dimensional grid cell is calculated as the estimated cost. The total cost is obtained based on the cumulative cost and the estimated cost. The D* algorithm is used to expand the three-dimensional grid cell with the minimum total cost, and the optimal path from the starting three-dimensional grid cell to the target three-dimensional grid cell is gradually retrieved and an optimal path sequence is generated.

[0121] Preferably, in this step, the operation area is divided into three-dimensional grid cells, and the risk area is defined in combination with the crack and corrosion areas. The cost function is set to comprehensively consider the risk distance and the height cost. The D* algorithm is used to dynamically optimize the operation route, ensure that the path avoids high-risk areas, reduce vertical movement, improve the safety and operability of path planning, and finally enable the operator to reach the target operation point efficiently and safely with the optimal path, effectively improving the operation efficiency and reducing risks, and ensuring the operation quality.

[0122] In summary, the beneficial effects of the tower feature matching method of the present invention are as follows: the point cloud data of the tower can be collected in real time by using the sensor carried by the unmanned aerial vehicle; through preprocessing and feature extraction, the accurate analysis of the structural state of the tower can be realized, including the tilt angle, cracks and corrosion areas; the random forest model is used to classify the tower features, and normal and abnormal samples can be effectively identified. By quantifying the prediction confidence, it can guide manual review. For abnormal samples, the D* algorithm is used to dynamically optimize the route of the operator to avoid high-risk areas. The present invention not only improves the automation level of tower detection, but also reduces manual intervention, improves the accuracy and operation safety of detection, and provides a strong guarantee for the stable operation of the power system.

[0123] Embodiment 2 is the second embodiment of the present invention. This embodiment provides a tower feature matching system, which includes a processing module for preprocessing the collected tower point cloud data and extracting the target point cloud data, where the target point cloud data is the point cloud data centered on the tower.

[0124] An analysis and detection module for performing structural analysis and feature detection of the pole tower on the target point cloud data and extracting pole tower features;

[0125] A classification module for classifying the pole tower features and dividing normal samples and abnormal samples;

[0126] An optimization module for dynamically planning the operation route of the operator according to the abnormal samples by using a path optimization algorithm.

[0127] Embodiment 3 is the third embodiment of the present invention. The difference from the previous two embodiments is that:

[0128] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that makes a contribution to the prior art, or a part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0129] The logic and / or steps represented in the flowchart or described in other ways herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch instructions from the instruction execution system, apparatus, or device and execute the instructions), or in combination with these instruction execution systems, apparatus, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transmit a program for use by or in combination with an instruction execution system, apparatus, or device.

[0130] More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection (electronic device) having one or more wirings, a portable computer diskette (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable media can even be paper or other suitable media on which a program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpretation, or other suitable processing as necessary, and then storing it in a computer memory.

[0131] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above-described embodiments, the multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having suitable combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0132] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A method for matching tower characteristics, characterized in that: including preprocessing the collected tower pole point cloud data to extract target point cloud data, where the target point cloud data is the point cloud data centered on the tower pole; conducting structural analysis and feature detection of the tower pole on the target point cloud data to extract tower pole features; classifying the tower pole features to divide normal samples and abnormal samples; using a path optimization algorithm to dynamically plan the operation route of the operator according to the abnormal samples.

2. The pole tower feature matching method according to claim 1, wherein: The preprocessing of the collected tower pole point cloud data includes: using a drone equipped with a sensor to collect tower pole point cloud data; conducting integrity inspection on the collected tower pole point cloud data to remove error data and abnormal points; conducting alignment processing on the collected tower pole point cloud data and converting it to a unified coordinate system.

3. The tower feature matching method according to claim 2, characterized in that: The structural analysis and feature detection of the tower pole on the target point cloud data include: dividing the target point cloud data into blocks of tower feet, tower bodies, and tower tops according to the physical structure of the tower pole; calculating the geometric center and tilt angle of each block; detecting crack and corrosion areas for each block.

4. The tower feature matching method according to claim 3, wherein: The classification of the tower pole features to divide normal samples and abnormal samples includes: normalizing the tower pole features according to the sample labels manually marked; using a random forest model to classify the tower pole features to divide normal samples and abnormal samples.

5. The pole and tower feature matching method according to claim 4, wherein: The using a path optimization algorithm to dynamically plan the operation route of the operator according to the abnormal samples includes: dividing the operation area into three-dimensional grid cells and defining risk areas; setting a cost function H for each three-dimensional grid cell, where the cost function H comprehensively considers the distance from the center of the three-dimensional grid cell to the risk area and the height cost; using the D* algorithm to dynamically optimize the operation route of the operator according to the cost function H.

6. The tower feature matching method according to claim 4, characterized in that: The method further includes evaluating the confidence level of the division result.

7. The tower feature matching method according to any one of claims 1 to 6, characterized in that: The tower pole features include but are not limited to tilt angle, number and size of cracks, and area of corrosion areas.

8. A system adopting the tower pole feature matching method as described in any one of claims 1 to 7, characterized in that: including a processing module for preprocessing the collected tower pole point cloud data to extract target point cloud data, where the target point cloud data is the point cloud data centered on the tower pole; an analysis and detection module for conducting structural analysis and feature detection of the tower pole on the target point cloud data to extract tower pole features; a classification module for classifying the tower pole features to divide normal samples and abnormal samples; an optimization module for using a path optimization algorithm to dynamically plan the operation route of the operator according to the abnormal samples.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

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

Images