A power transmission point cloud classification and deployment method, system, device and storage medium
By collecting visible light images and point cloud data using drones, and combining the improved YOLOv8 segmentation algorithm and BRA attention mechanism, the real-time performance and efficiency issues of point cloud classification in power grid inspection were resolved. This enabled efficient and accurate identification of transmission lines, meeting the needs of real-time power grid inspection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANDONG ZHIYANG ELECTRIC
- Filing Date
- 2026-01-19
- Publication Date
- 2026-06-05
AI Technical Summary
Existing point cloud classification algorithms struggle to meet the requirements of real-time performance and inspection efficiency in power grid inspections, especially when faced with a significant increase in point cloud data volume, where class imbalance and the balance between real-time performance and efficiency are difficult to resolve.
Visible light images and point cloud data of power transmission lines were collected by drones. The high resolution and rich texture information of the visible light images were used to train the segmentation model. Combined with the improved YOLOv8 segmentation algorithm and BRA attention mechanism, the data was mapped and converted into point cloud data classification results. Euclidean distance clustering was used to optimize the unclassified point cloud data to achieve efficient classification.
It improves the accuracy and efficiency of point cloud classification, meets the field application needs of real-time power grid inspection, solves the problems of slow data acquisition, poor data quality and incomplete classification in traditional methods, and ensures the reliability and completeness of classification results.
Smart Images
Figure CN122157030A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of computer vision technology, and in particular to a method, system, device and storage medium for classifying and deploying power transmission point clouds. Background Technology
[0002] As power grids expand in scale and become more complex, traditional inspection methods are inefficient and unable to meet high-precision requirements. Airborne lidar technology can quickly acquire three-dimensional point cloud data of transmission channels, providing an efficient means for intelligent inspection.
[0003] Transmission line point cloud classification is a key technology in power line inspection and safety management. It is primarily used to extract key elements such as transmission lines, towers, vegetation, and buildings from high-density 3D point cloud data acquired by airborne lidar. However, traditional point cloud classification algorithms struggle to meet the requirements of real-time performance and inspection efficiency. In recent years, deep learning-based point cloud classification methods have gradually become mainstream. Nevertheless, with the significant increase in point cloud data volume, class imbalance and the balance between real-time performance and efficiency limit its application in the field. Summary of the Invention
[0004] This application provides a method, system, device, and storage medium for power transmission point cloud classification and deployment, which solves the technical problem that existing point cloud classification algorithms cannot meet the requirements of real-time performance and inspection efficiency in power grid inspection.
[0005] This invention utilizes unmanned aerial vehicles (UAVs) to collect visible light images and point cloud data of power transmission lines during inspections. The high resolution and rich texture information of the visible light images are used to train a segmentation model. Based on the segmentation results from the visible light images, the data is mapped and converted into corresponding point cloud data classification results, thereby achieving efficient point cloud classification and recognition. This method not only meets the real-time and efficiency requirements of point cloud classification but also further improves the accuracy of point cloud classification by constructing an improved YOLOv8 segmentation algorithm model, providing efficient technical support for intelligent inspection and safety management of power transmission lines. The specific invention content of this application is described in detail below.
[0006] First, this application provides a method for classifying and deploying power transmission point clouds, the method comprising the following steps: Step S1: Collect visible light and point cloud data of the power transmission scene using drones, and obtain training dataset, test dataset and validation dataset after preprocessing; Step S2: Construct an improved YOLOv8 segmentation algorithm model. The improved YOLOv8 algorithm includes: replacing the loss function with DiceLoss loss to make the model pay more attention to foreground information, and introducing the BRA attention mechanism to further improve the recognition accuracy of edge segmentation. Step S3: Based on the improved YOLOv8 segmentation algorithm model, the model is trained on visible light data using the training dataset to obtain the segmentation mask image of the data to be identified, the corresponding point cloud data is converted into a depth map, the depth map is processed using the obtained segmentation mask image, and the point cloud classification result is obtained. Step S4: Based on the point cloud data classification results, use a clustering algorithm based on Euclidean distance to optimize the point cloud classification results, and then deploy the optimized point cloud classification results.
[0007] In one implementation of this application, step S1, which involves obtaining a training dataset, a test dataset, and a validation dataset after preprocessing, specifically involves using a labeling tool to segment and label visible light data to create a segmentation dataset.
[0008] In one implementation of this application, step S2, which involves constructing an improved YOLOv8 segmentation algorithm model, specifically includes: The loss function is replaced with the DiceLoss loss function, which makes the model focus more on extracting foreground information. The DiceLoss loss function is as follows: , in, This represents the actual value of pixel i. Let N represent the predicted value of pixel i, and N represent the total number of pixels.
[0009] In one implementation of this application, step S2, the BRA attention mechanism, specifically includes: The input feature data is divided into Several different regions, each region containing There are eigenvectors, where... , The feature data consists of height and width, which are then linearly mapped to obtain three inputs. , and ,right , Find the average value. , ,pass , Construct the adjacency matrix vector of regional correlation: In the formula, To construct the adjacency matrix vector, Represents the transpose of a matrix; Preserve the front of each area The index of the connections In the formula, For the index matrix, clustering , , , After aggregation Using attention operations: In the formula, Let be the attention function. Parameterization is achieved using depthwise separable convolutions. It is a normalized exponential function.
[0010] In one implementation of this application, step S3 specifically includes: The segmentation model is trained based on the improved YOLOv8 segmentation algorithm and the training dataset of annotated visible light images to obtain a semantic segmentation model of the improved YOLOv8 segmentation algorithm for visible light images. Based on the improved YOLOv8 segmentation algorithm model, a semantic segmentation mask map of the power transmission region of the visible light image to be identified is obtained. Obtain the point cloud data corresponding to the visible light image to be identified, convert the point cloud data into a depth map, map the visible light image segmentation mask to the point cloud depth map, obtain the classification result of the point cloud depth map, convert the classified depth map back into point cloud data, and obtain the classification result of the point cloud data.
[0011] In one implementation of this application, step S4 specifically includes: Based on the point cloud classification results, the classified points are labeled with category labels {c1, c2}, and the center point of each category is obtained. , To belong to category All point sets; Obtain the unclassified point cloud from the raw point cloud data, and for each unclassified point... For unclassified points, a clustering algorithm based on Euclidean distance is used to calculate their Euclidean distances to the centers of all classes: ,Will Assign to the nearest category: Optimize point cloud classification results.
[0012] In one implementation of this application, step S4, which involves deploying the optimized point cloud classification results, specifically involves the following process: Prepare the deployment environment using the Ubuntu system, download the Rockchip cross-compilation toolchain, add the cross-compilation toolchain path to the environment variables, and install the RKNN Toolkit and its dependencies. The improved YOLOv8 segmentation model was converted into an ONNX model, and the installed RKNN Toolkit was used to convert the ONNX model into an RKNN model. Use the example code provided in the RKNN Model Zoo to compile the code and generate an executable file; Upload the RKNN model and executable file to the Rockchip development board, run the executable file for inference testing, and deploy and test the algorithm on the Rockchip development board.
[0013] Secondly, this application also provides a power transmission point cloud classification and deployment system, the system comprising: The preprocessing module is used to collect visible light and point cloud data of power transmission scenarios by drones, and then preprocess the data to obtain training datasets, test datasets and validation datasets. The model building module is used to build an improved YOLOv8 segmentation algorithm model. The improvements to the YOLOv8 algorithm include: replacing the loss function with DiceLoss loss to make the model pay more attention to foreground information, and introducing the BRA attention mechanism to further improve the recognition accuracy of edge segmentation. The model training module is used to train the model based on the improved YOLOv8 segmentation algorithm model using the training dataset for visible light data, obtain the segmentation mask map of the data to be identified, convert the corresponding point cloud data into a depth map, process the depth map using the obtained segmentation mask map, and obtain the point cloud classification result. The optimization deployment module is used to optimize the point cloud classification results based on the point cloud data using a clustering algorithm based on Euclidean distance, and then deploy the optimized point cloud classification results.
[0014] Secondly, this application also provides a power transmission point cloud classification and deployment device, the device comprising: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which enables the at least one processor to perform the aforementioned method for classifying and deploying power transmission point clouds.
[0015] Finally, this application also provides a non-volatile computer storage medium for power transmission point cloud classification and deployment, which stores computer-executable instructions that are executed by a processor to implement the aforementioned power transmission point cloud classification and deployment method.
[0016] The power transmission point cloud classification and deployment method, system, equipment, and storage medium provided in this application have the following beneficial effects: (1) By efficiently collecting visible light and point cloud data of power transmission scenarios through drones, and combining them with annotation tools to create segmented datasets and divide them into training, testing and validation sets, the problem of slow and poor quality data acquisition in traditional inspections is solved. This provides a multi-dimensional and high-quality data foundation for model training, avoids classification bias caused by insufficient or poor data, and ensures the reliability of subsequent processes.
[0017] (2) The YOLOv8 loss function is replaced with DiceLoss to focus on foreground information. The BRA attention mechanism is introduced to optimize regional correlation and edge recognition, effectively alleviate the problem of class imbalance in power transmission scenarios, enhance the edge segmentation capability of key elements such as towers and conductors, capture power transmission area features more accurately than traditional models, and improve the accuracy of semantic segmentation.
[0018] (3) The visible light segmentation mask image is used to process the point cloud depth map, and the visible light high texture and point cloud three-dimensional information are integrated to avoid the limitation of traditional point cloud classification relying on single data. This achieves accurate mapping from image segmentation to point cloud classification, which reduces the complexity of direct point cloud processing and ensures the integrity of classification, thus greatly improving the efficiency of power transmission point cloud classification.
[0019] (4) The unclassified point cloud is optimized by Euclidean distance clustering to supplement the classification gap; the model is deployed by relying on Rockchip toolchain and RKNNToolkit, which solves the problem that the traditional algorithm classification is incomplete and difficult to implement on site, and ensures that the optimized classification results can run efficiently on the inspection equipment to meet the field application needs of real-time power grid inspection. Attached Figure Description
[0020] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart illustrating a power transmission point cloud classification and deployment method provided in this application embodiment; Figure 2 A schematic diagram of the improved YOLOv8 network structure provided in the embodiments of this application; Figure 3 This is a schematic diagram of the BRA attention module structure provided in an embodiment of this application; Figure 4 A flowchart of power transmission point cloud classification based on segmentation algorithm provided for embodiments of this application; Figure 5 This is a depth result image of visible light image segmentation and point cloud classification provided in the embodiments of this application; Figure 6 A deployment flowchart provided for an embodiment of this application; Figure 7A diagram illustrating the composition of a power transmission point cloud classification and deployment system provided in this application embodiment; Figure 8 This is a schematic diagram of a power transmission point cloud classification and deployment device provided in an embodiment of this application. Detailed Implementation
[0021] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0022] This application provides a method, system, device, and storage medium for power transmission point cloud classification and deployment. The technical solutions proposed in this application will be described in detail below with reference to the accompanying drawings.
[0023] Figure 1 This is a flowchart illustrating a power transmission point cloud classification and deployment method provided in an embodiment of this application. Figure 1 As shown, the method mainly includes the following steps: Step S1: Collect visible light and point cloud data of the power transmission scene using drones, and obtain training dataset, test dataset and validation dataset after preprocessing.
[0024] Specifically, visible light and point cloud data of power transmission scenarios are acquired by drones. The visible light and point cloud data of power transmission scenarios are collected by drones, and the Labelme annotation tool is used to segment and annotate different categories of visible light data to create a segmentation dataset. The categories include: towers, conductors, etc. The corresponding JSON format annotation files are obtained. The annotated visible light segmentation dataset is randomly divided into training set, validation set and test set in a ratio of 7:2:1.
[0025] Step S2: Construct an improved YOLOv8 segmentation algorithm model. The improved YOLOv8 algorithm includes: replacing the loss function with DiceLoss loss to make the model pay more attention to foreground information, and introducing the BRA attention mechanism to further improve the recognition accuracy of edge segmentation.
[0026] Specifically, improving the YOLOv8 algorithm network model, such as Figure 2 As shown. In YOLOv8, the loss function is replaced with the DiceLoss loss function, making the model focus more on extracting foreground information. The DiceLoss loss function is: ,in, This represents the actual value of pixel i. Let N represent the predicted value of pixel i, and N represent the total number of pixels.
[0027] BRA attention mechanism, such as Figure 3 As shown, it specifically includes: The input feature data is divided into Several different regions, each region containing There are eigenvectors, where... , The feature data consists of height and width, which are then linearly mapped to obtain three inputs. , and ,right , Find the average value. , ,pass , Construct the adjacency matrix vector of regional correlation: In the formula, To construct the adjacency matrix vector, Represents the transpose of a matrix; Preserve the front of each area The index of the connections In the formula, For the index matrix, clustering , , , After aggregation Using attention operations: In the formula, Let be the attention function. Parameterization is achieved using depthwise separable convolutions. It is a normalized exponential function.
[0028] Step S3: Based on the improved YOLOv8 segmentation algorithm model, the model is trained on visible light data using the training dataset to obtain the segmentation mask image of the data to be identified, the corresponding point cloud data is converted into a depth map, the depth map is processed using the obtained segmentation mask image, and the point cloud classification result is obtained.
[0029] Specifically, based on the improved YOLOv8 segmentation algorithm model, model training is performed on visible light data to obtain a segmentation mask image of the data to be identified. The corresponding point cloud data is then converted into a depth map. The depth map is processed using the obtained segmentation mask image to obtain the point cloud classification result. The specific process is as follows: Figure 4 As shown.
[0030] During the implementation process, based on the improved YOLOv8 segmentation algorithm, the segmentation model was trained on the training dataset of labeled visible light images to obtain a semantic segmentation model of the transmission area of the visible light image using the improved YOLOv8 algorithm. Based on the improved YOLOv8 segmentation algorithm model, a semantic segmentation mask map of the power transmission region in the visible light image to be identified is obtained, such as... Figure 5 As shown: the area with a pixel value of 255 represents the background area, and the areas with pixel values of 0 and 1 represent the power transmission scene area.
[0031] In step S3, the corresponding point cloud data are converted into depth maps, and the depth maps are processed using the obtained segmentation mask image to obtain the point cloud classification results, including: Obtain the point cloud data corresponding to the visible light image to be identified, load camera parameters to obtain the projection transformation matrix M, rotation matrix R, and translation matrix T of the point cloud data, and use... Convert point cloud data from world coordinates to camera coordinates. Point cloud in world coordinate system For the point cloud in the camera coordinate system, extract the depth value of each point in the camera coordinate system, and use M to project it onto the image plane to convert the point cloud data into a point cloud depth map. The segmentation mask map of the visible light image to be identified is mapped onto the point cloud depth map, and the classification result of the point cloud depth map for each category is obtained as follows: Figure 5 As shown; For each category of point cloud depth map, the image planar points are projected onto a 3D plane using M. Convert to point cloud data, merge single-category point cloud data, and obtain the classification results of the point cloud data.
[0032] Step S4: Based on the point cloud data classification results, use a clustering algorithm based on Euclidean distance to optimize the point cloud classification results, and then deploy the optimized point cloud classification results.
[0033] In this application, based on the point cloud data classification results obtained in step S3, the classified points are labeled with category labels {c1, c2}, and the center point of each category is obtained. , To belong to category All point sets; Obtain the unclassified point cloud from the raw point cloud data, and for each unclassified point... For unclassified points, a clustering algorithm based on Euclidean distance is used to calculate their Euclidean distances to the centers of all classes: ,Will Assign to the nearest category: This can further optimize the point cloud classification results.
[0034] Furthermore, the deployment process of the power transmission point cloud classification method based on segmentation algorithms is as follows: Figure 6 As shown, it specifically includes: Prepare the deployment environment using the Ubuntu system. Download the Rockchip cross-compilation toolchain, add the cross-compilation toolchain path to the environment variables, and install the RKNN Toolkit and its dependencies.
[0035] The model conversion process involves converting the improved YOLOv8 segmentation model into an ONNX model, and then using the installed RKNNToolkit to convert the ONNX model into a RKNN model.
[0036] The code is compiled using the example code provided in the RKNN Model Zoo to generate an executable file.
[0037] For inference testing, the RKNN model and executable file are uploaded to the Rockchip development board. The executable file is run to perform inference testing, and the algorithm is deployed and tested on the Rockchip development board. The above describes a method for classifying and deploying power transmission point clouds according to embodiments of this application. Based on the same inventive concept, embodiments of this application also provide a system for classifying and deploying power transmission point clouds. Figure 7 A diagram illustrating the composition of a power transmission point cloud classification and deployment system provided in this application embodiment is shown below. Figure 7 As shown, the system mainly includes: a preprocessing module 701, which is used to collect visible light and point cloud data of power transmission scenarios through UAVs, and obtain training datasets, test datasets and verification datasets after preprocessing; The model building module 702 is used to build an improved YOLOv8 segmentation algorithm model. The improved YOLOv8 algorithm includes: replacing the loss function with DiceLoss loss to make the model pay more attention to foreground information, and introducing the BRA attention mechanism to further improve the recognition accuracy of edge segmentation. The model training module 703 is used to train the model based on the improved YOLOv8 segmentation algorithm model using the training dataset to obtain the segmentation mask map of the data to be identified, convert the corresponding point cloud data into a depth map, process the depth map using the obtained segmentation mask map, and obtain the point cloud classification result. The optimization deployment module 704 is used to optimize the point cloud classification results based on the point cloud data classification results using a clustering algorithm based on Euclidean distance, and then deploy the optimized point cloud classification results.
[0038] The above describes a power transmission point cloud classification and deployment system provided in this application. Based on the same inventive concept, this application also provides a power transmission point cloud classification and deployment device. Figure 8This application provides a schematic diagram of a power transmission point cloud classification and deployment device, as shown in the embodiments. Figure 8 As shown, the device mainly includes: at least one processor 801; and a memory 802 communicatively connected to the at least one processor; wherein the memory 802 stores instructions that can be executed by the at least one processor 801, and the instructions are executed by the at least one processor 801 to enable the at least one processor 801 to complete the aforementioned power transmission point cloud classification and deployment method.
[0039] In addition, embodiments of this application also provide a non-volatile computer storage medium for power transmission point cloud classification and deployment, which stores computer-executable instructions that are executed by a processor to implement the aforementioned power transmission point cloud classification and deployment method.
[0040] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0041] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0042] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0043] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0044] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.
[0045] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0046] The above description is merely an embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of this application should be included within the scope of the claims of this application.
Claims
1. A method for classifying and deploying power transmission point clouds, characterized in that, The method includes the following steps: Step S1: Collect visible light and point cloud data of the power transmission scene using drones, and obtain training dataset, test dataset and validation dataset after preprocessing; Step S2: Construct an improved YOLOv8 segmentation algorithm model. The improved YOLOv8 algorithm includes: replacing the loss function with DiceLoss loss to make the model pay more attention to foreground information, and introducing the BRA attention mechanism to further improve the recognition accuracy of edge segmentation. Step S3: Based on the improved YOLOv8 segmentation algorithm model, the model is trained on visible light data using the training dataset to obtain the segmentation mask image of the data to be identified, the corresponding point cloud data is converted into a depth map, the depth map is processed using the obtained segmentation mask image, and the point cloud classification result is obtained. Step S4: Based on the point cloud data classification results, use a clustering algorithm based on Euclidean distance to optimize the point cloud classification results, and then deploy the optimized point cloud classification results.
2. The method for classifying and deploying power transmission point clouds according to claim 1, characterized in that, In step S1, after preprocessing, training dataset, test dataset, and validation dataset are obtained. Specifically, visible light data is segmented and labeled using an annotation tool to create a segmentation dataset.
3. The method for classifying and deploying power transmission point clouds according to claim 1, characterized in that, In step S2, an improved YOLOv8 segmentation algorithm model is constructed, specifically as follows: The loss function is replaced with the DiceLoss loss function, which makes the model focus more on extracting foreground information. The DiceLoss loss function is as follows: , in, This represents the actual value of pixel i. Let N represent the predicted value of pixel i, and N represent the total number of pixels.
4. The method for classifying and deploying power transmission point clouds according to claim 1, characterized in that, In step S2, the BRA attention mechanism specifically includes: The input feature data is divided into Several different regions, each region containing There are eigenvectors, where... , The feature data consists of height and width, which are then linearly mapped to obtain three inputs. , and ,right , Find the average value. , ,pass , Construct the adjacency matrix vector of regional correlation: In the formula, To construct the adjacency matrix vector, Represents the transpose of a matrix; Preserve the front of each area The index of the connections In the formula, For the index matrix, clustering , , , After aggregation Using attention operations: In the formula, Let be the attention function. Parameterization is achieved using depthwise separable convolutions. It is a normalized exponential function.
5. The method for classifying and deploying power transmission point clouds according to claim 1, characterized in that, Step S3 specifically includes: The segmentation model is trained based on the improved YOLOv8 segmentation algorithm and the training dataset of annotated visible light images to obtain a semantic segmentation model of the improved YOLOv8 segmentation algorithm for visible light images. Based on the improved YOLOv8 segmentation algorithm model, a semantic segmentation mask map of the power transmission region of the visible light image to be identified is obtained. Obtain the point cloud data corresponding to the visible light image to be identified, convert the point cloud data into a depth map, map the visible light image segmentation mask to the point cloud depth map, obtain the classification result of the point cloud depth map, convert the classified depth map back into point cloud data, and obtain the classification result of the point cloud data.
6. The method for classifying and deploying power transmission point clouds according to claim 1, characterized in that, Step S4 specifically includes: Based on the point cloud classification results, the classified points are labeled with category labels {c1, c2}, and the center point of each category is obtained. , To belong to category All point sets; Obtain the unclassified point cloud from the raw point cloud data, and for each unclassified point... For unclassified points, a clustering algorithm based on Euclidean distance is used to calculate their Euclidean distances to the centers of all classes: ,Will Assign to the nearest category: Optimize point cloud classification results.
7. The method for classifying and deploying power transmission point clouds according to claim 1, characterized in that, In step S4, the optimized point cloud classification results are deployed, and the specific process is as follows: Prepare the deployment environment using the Ubuntu system, download the Rockchip cross-compilation toolchain, add the cross-compilation toolchain path to the environment variables, and install the RKNN Toolkit and its dependencies. The improved YOLOv8 segmentation model was converted into an ONNX model, and the installed RKNN Toolkit was used to convert the ONNX model into an RKNN model. Use the example code provided in the RKNN Model Zoo to compile the code and generate an executable file; Upload the RKNN model and executable file to the Rockchip development board, run the executable file for inference testing, and deploy and test the algorithm on the Rockchip development board.
8. A power transmission point cloud classification and deployment system, characterized in that, The system includes: The preprocessing module is used to collect visible light and point cloud data of power transmission scenarios by drones, and then preprocess the data to obtain training datasets, test datasets and validation datasets. The model building module is used to build an improved YOLOv8 segmentation algorithm model. The improvements to the YOLOv8 algorithm include: replacing the loss function with DiceLoss loss to make the model pay more attention to foreground information, and introducing the BRA attention mechanism to further improve the recognition accuracy of edge segmentation. The model training module is used to train the model based on the improved YOLOv8 segmentation algorithm model using the training dataset for visible light data, obtain the segmentation mask map of the data to be identified, convert the corresponding point cloud data into a depth map, process the depth map using the obtained segmentation mask map, and obtain the point cloud classification result. The optimization deployment module is used to optimize the point cloud classification results based on the point cloud data using a clustering algorithm based on Euclidean distance, and then deploy the optimized point cloud classification results.
9. A power transmission point cloud classification and deployment device, characterized in that, The device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform a power transmission point cloud classification and deployment method according to any one of claims 1-7.
10. A non-volatile computer storage medium for power transmission point cloud classification and deployment, storing computer-executable instructions, characterized in that, The computer-executable instructions are executed by the processor to implement the power transmission point cloud classification and deployment method according to any one of claims 1-7.