CN117635630A - Automatic object segmentation method and system based on substation scenic spot cloud - Google Patents

Abstract

The invention discloses an automatic object segmentation method and system based on real point clouds of substations. The invention uses an adaptive layering method to dynamically adjust and sample according to the distribution and characteristics of point cloud data, so that each level has an appropriate number and distribution of points. It can capture the multi-scale features of point cloud data and effectively represent it from global to local. This is particularly important when dealing with scale differences and multi-scale information in point cloud data; Pointnet++ can extract features on each local area and use hierarchical aggregation operations to obtain more global context information. The network can make full use of local and global Information for more accurate segmentation tasks, Pointnet++ can operate more efficiently on each local region, reducing computational and memory requirements. At the same time, through the hierarchical structure, the level of segmentation operations can be selected in a targeted manner to further improve calculation efficiency.

Description

Automatic object segmentation method and system based on real point cloud of substation

Technical field

The invention relates to the technical fields of computer vision and image processing, and in particular to an automatic object segmentation method and system based on real point clouds of substations.

Background technique

After years of development, three-dimensional vision technology has become more mature and has been widely used in industrial production. In order to improve the intelligence of substation management, three-dimensional technology can be used to comprehensively display the data, images and videos of the substation. When performing three-dimensional modeling, due to the large and complex substation scene, in order to ensure the accuracy and efficiency of model acquisition, laser scanning can be used to obtain point cloud data. Currently used point cloud segmentation methods include graph-based methods: point cloud data is modeled as a graph structure, in which the points of the point cloud are used as nodes of the graph, and segmentation is performed by calculating the edges and edge characteristics between nodes; based on Regional methods: Divide point cloud data into continuous regions, and use clustering or segmentation algorithms to mark regions; Deep learning-based methods: This method uses deep learning, such as convolutional neural networks and graph convolutional neural networks. Network and attention mechanisms are used to segment point cloud data; feature-based methods: segmentation is performed by extracting features in point cloud data, such as shape descriptors, normals, curvature, and color. Then only clustering or other algorithms are used to analyze and segment the features. Method based on set features: Use set features, such as the position, distance, normal direction, etc. of the point cloud, to segment the point cloud data. For example, segmentation methods based on datum planes, segmentation methods based on curvature, etc.

The above methods largely rely on the distribution of objects or the selection of initial seed points, so they are not robust. At the same time, the point cloud file contains a large number of points, including many useless noise points. Therefore, it is necessary to cluster and segment the point cloud to extract effective points for the reconstruction of the three-dimensional model.

Contents of the invention

In view of the above-mentioned existing problems, the present invention is proposed. Therefore, the present invention provides a method and system for automatic object segmentation based on real point clouds of substations, which solves the problem that the current traditional methods largely rely on the distribution of objects or the selection of initial seed points, and therefore are not robust; At the same time, point cloud files contain a large number of points, among which there are many useless noise points, making it impossible to extract effective points.

In order to solve the above technical problems, the present invention provides the following technical solutions:

In the first aspect, the present invention provides an automatic object segmentation method based on real point clouds of substations, including:

Obtain real-life point cloud data and corresponding annotation information;

The real-life point cloud data is preprocessed through an adaptive layering method. The preprocessing divides the point cloud data into different levels and arranges the organization of points;

Perform feature extraction for each level of point cloud, and fuse point cloud features at different levels to obtain global and comprehensive feature representation;

Use a fully connected layer to predict the category of each point, segment and semantically label the point cloud based on the prediction results, and perform training and optimization;

Use the trained network to segment the new real-life point cloud and obtain the semantic label of each point to segment, extract and analyze different objects based on the label.

As a preferred solution of the object automatic segmentation method based on the real-life point cloud of a substation according to the present invention, the obtaining of the real-life point cloud data and the corresponding annotation information, wherein the obtaining of the real-life point cloud data includes:

Obtain information on the geometry and location of buildings, equipment, power lines, and heat distribution;

The corresponding annotation information includes category annotation, bounding box annotation, and point cloud annotation.

As a preferred solution of the object automatic segmentation method based on the real-life point cloud of a substation according to the present invention, the real-life point cloud data is preprocessed through an adaptive layering method, and the preprocessing divides the point cloud data into Organizational methods for different levels and arrangement of points, including,

Through the octree method, a hierarchical structure is constructed according to the distribution and characteristics of the point cloud to obtain point clouds of different levels, including setting four levels;

The first level includes cloud data of the entire substation, providing global characteristics of the substation;

The second level divides the substation into different heat distribution areas based on the point cloud data of heat distribution;

The third level divides substation site cloud data into different equipment functional areas based on the distribution and complexity;

The fourth level divides the grid according to the type and function combination of the equipment inside each area.

As a preferred solution of the object automatic segmentation method based on real-life point clouds of substations according to the present invention, the arrangement of points is used to obtain point clouds of different levels with different densities and resolutions. Specifically, include,

In the first level and the second level, the first neighborhood radius threshold is set through the local neighborhood point cloud method to determine the corresponding local neighborhood;

In the third level and the fourth level, the second neighborhood radius threshold is set through the local neighborhood point cloud method to determine the corresponding local neighborhood;

The first neighborhood radius threshold is greater than the second neighborhood radius threshold.

As a preferred solution of the object automatic segmentation method based on the real point cloud of the substation according to the present invention, it also includes: after determining the local neighborhood, setting the resolution according to the relative position of the point or projection to different grids. The resolution of each level;

The first level and the second level set the first resolution;

The third and fourth levels set the second resolution;

The first resolution is smaller than the second resolution;

In the hierarchical structure, for each level of point cloud data, the local coordinates relative to that level are calculated for use in the segmentation stage.

As a preferred solution of the object automatic segmentation method based on the real point cloud of the substation according to the present invention, feature extraction is performed for each level of point cloud, and the point cloud features of different levels are fused to obtain the global and comprehensive The characteristic representation includes,

At each level after adaptive layering, feature extraction is performed on local point cloud data through the Pointnet++ grid structure, including multi-layer perception and feature extraction of local point clouds;

Use a fully connected layer to predict the category of each point. Based on the prediction results, the point cloud is segmented and semantically labeled, and trained and optimized, including,

For specific point cloud segmentation tasks, the segmentation head network is introduced in Pointnet++ to output the semantic segmentation label of each point, and the segmentation head is adjusted according to the task requirements;

In the training phase, point cloud data with known labels are used for supervised learning, and the parameters of the network are updated through backpropagation and optimization algorithms, with the goal of minimizing the loss function between the predicted labels and the true labels.

As a preferred solution of the object automatic segmentation method based on the real-life point cloud of the substation according to the present invention, the new real-life point cloud is segmented using the trained network to obtain the semantic label of each point, so as to classify the new real-life point cloud according to the label. Segment, extract and analyze different objects including,

In the testing phase, the trained Pointnet++ network model is used to segment the new point cloud data, forward propagation is performed on each point, and the corresponding label is assigned to the point based on the final segmentation result.

In the second aspect, the present invention provides a system for automatic segmentation of objects based on real-life point clouds of substations, including:

The acquisition module is used to obtain real-life point cloud data and corresponding annotation information;

A preprocessing module for preprocessing the real-life point cloud data through an adaptive layering method. The preprocessing divides the point cloud data into different levels and arranges the organization of points;

The fusion module is used to extract features from point clouds at each level and fuse point cloud features at different levels to obtain global and comprehensive feature representation;

The training module is used to predict the category of each point using a fully connected layer, segment and semantically label the point cloud based on the prediction results, and perform training and optimization;

The segmentation module is used to use the trained network to segment new real-life point clouds and obtain the semantic labels of each point to segment, extract and analyze different objects based on the labels.

In a third aspect, the present invention provides a computing device, including:

memory and processor;

The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the object automatic segmentation method based on the real point cloud of the substation are implemented.

In a fourth aspect, the present invention provides a computer-readable storage medium that stores computer-executable instructions. When the computer-executable instructions are executed by a processor, the steps of the object automatic segmentation method based on real point clouds of substations are implemented.

Compared with the existing technology, the beneficial effects of the present invention are: the present invention uses an adaptive layering method to dynamically adjust and sample according to the distribution and characteristics of point cloud data, so that each level has an appropriate number and distribution of points. It can capture the multi-scale features of point cloud data and effectively represent it from global to local. This is particularly important when dealing with scale differences and multi-scale information in point cloud data; Pointnet++ can extract features on each local area and use hierarchical aggregation operations to obtain more global context information. The network can make full use of local and global Information for more accurate segmentation tasks, Pointnet++ can operate more efficiently on each local region, reducing computational and memory requirements. At the same time, through the hierarchical structure, the level of segmentation operations can be selected in a targeted manner to further improve calculation efficiency.

Description of drawings

In order to explain the technical solutions of the embodiments of the present invention more clearly, the drawings needed to be used in 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. Those of ordinary skill in the art can also obtain other drawings based on these drawings without exerting any creative effort. in:

Figure 1 is a schematic flowchart of the overall object automatic segmentation method based on a real point cloud of a substation according to an embodiment of the present invention.

Detailed ways

In order to make the above objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings. It is obvious that the described embodiments are part of the embodiments of the present invention, not all of them. Example. Based on the embodiments of the present invention, all other embodiments obtained by ordinary people in the art without creative efforts should fall within the protection scope of the present invention.

Many specific details are set forth in the following description to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described here. Those skilled in the art can do so without departing from the connotation of the present invention. Similar generalizations are made, and therefore the present invention is not limited to the specific embodiments disclosed below.

Second, reference herein to "one embodiment" or "an embodiment" refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. "In one embodiment" appearing in different places in this specification does not all refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.

The present invention will be described in detail with reference to schematic diagrams. When describing the embodiments of the present invention in detail, for the convenience of explanation, the cross-sectional diagrams showing the device structure will be partially enlarged according to the general scale. Moreover, the schematic diagrams are only examples and shall not limit the present invention. scope of protection. In addition, the three-dimensional dimensions of length, width and depth should be included in actual production.

At the same time, in the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "upper, lower, inner and outer" are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention. The invention and simplified description are not intended to indicate or imply that the systems or elements referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore are not to be construed as limitations of the invention. Furthermore, the terms "first, second or third" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance.

Unless otherwise clearly stated and limited in the present invention, the terms "installation, connection, and connection" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integrated connection; it can also be a mechanical connection, an electrical connection, or a direct connection. A connection can also be indirectly connected through an intermediary, or it can be an internal connection between two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood on a case-by-case basis.

Example 1

Referring to Figure 1, an embodiment of the present invention provides an automatic object segmentation method based on real point clouds of substations, including:

S1: Obtain real-life point cloud data and corresponding annotation information;

Furthermore, obtaining real-life point cloud data and corresponding annotation information, wherein obtaining real-life point cloud data includes:

Obtain information on the geometry and location of buildings, equipment, power lines, and heat distribution;

The corresponding annotation information includes category annotation, bounding box annotation, and point cloud annotation.

It should be noted that data can be obtained through laser scanners, thermal imaging cameras, etc. Laser scanners can measure the distance and position of objects and generate point cloud data. The laser scanner can obtain three-dimensional point cloud data of various parts of the substation; the thermal imager can measure the infrared radiation of the object and generate a heat map. Thermal image data can be obtained from various parts of the substation through thermal imaging cameras, ensuring that the data set contains enough samples to represent all aspects of the substation scene. At the same time, annotating the collected real-life data can provide more semantic information and scene understanding.

S2: Preprocess the real-life point cloud data through the adaptive layering method. The preprocessing divides the point cloud data into different levels and arranges the organization of points;

Furthermore, the real-life point cloud data is preprocessed through the adaptive layering method. The preprocessing divides the point cloud data into different levels and arranges the organization of the points, including,

Through the octree method, a hierarchical structure is constructed according to the distribution and characteristics of the point cloud to obtain point clouds of different levels, including setting four levels;

The first level includes cloud data of the entire substation, providing global characteristics of the substation;

The second level divides the substation into different heat distribution areas based on the point cloud data of heat distribution;

The third level divides substation site cloud data into different equipment functional areas based on the distribution and complexity;

The fourth level divides the grid according to the type and function combination of the equipment inside each area.

It should be noted that for the substation scenario, considering the amount of calculation and level complexity, it is preferable to establish four levels for construction. Each level can perform subsequent local neighborhood settings according to its own characteristics. At the same time, for each level, the octree method is used to dynamically adjust and sample according to the density and characteristics of the point cloud, which can ensure that each level has an appropriate number and distribution of points, for example, subsampling or subsampling according to the density of points. segmentation.

It should also be noted that the purpose of adaptive layering is to divide the substation site cloud data into multiple levels according to the type of real scene, so that the point clouds at different levels have different densities and resolutions. The purpose of adaptive layering is to capture Multi-scale features of point cloud data.

Furthermore, the point organization method is used to obtain point clouds of different levels with different densities and resolutions, including:

In the first level and the second level, the first neighborhood radius threshold is set through the local neighborhood point cloud method to determine the corresponding local neighborhood;

In the third and fourth levels, the second neighborhood radius threshold is set through the local neighborhood point cloud method to determine the corresponding local neighborhood;

The first neighborhood radius threshold is greater than the second neighborhood radius threshold.

It should be noted that the local neighborhood point cloud method can be applied to multi-equipment scenarios in substations. It can retain the local structure and shape of the point cloud, which is conducive to detecting and analyzing details in specific areas. It is also useful for detecting specific equipment or performing semantic segmentation. For tasks, local neighborhood point clouds are more suitable.

At the same time, in order to reduce the amount of calculation, only two radius thresholds are set between levels to ensure appropriate density and at the same time, the suitability of the calculation amount is ensured according to different area divisions and the calculation efficiency is improved.

Furthermore, it also includes, after determining the local neighborhood, setting the resolution of each level according to the relative position of the point or the resolution of projection to different grids;

The first level and the second level set the first resolution;

The third and fourth levels set the second resolution;

The first resolution is smaller than the second resolution;

In the hierarchical structure, for each level of point cloud data, the local coordinates relative to that level are calculated for use in the segmentation stage.

It should be noted that setting different resolutions for different levels of recognition and segmentation requirements will help preserve local structure and morphological information in more detail and improve the corresponding computing efficiency. The step of determining levels based on relative positions or resolutions projected onto different grids helps determine the segmentation of point clouds and division of local areas, while the organization of local neighborhood point clouds further processes the point cloud data at each level to Preserve local features and extract detailed information.

For resolution, the first resolution can be set to 1 meter/div, and the second resolution can be set to 0.2 meters/div.

S3: Extract features from point clouds at each level, and fuse point cloud features at different levels to obtain global and comprehensive feature representation;

Furthermore, feature extraction is performed for each level of point cloud, and point cloud features at different levels are fused to obtain global and comprehensive feature representation, including,

At each level after adaptive layering, feature extraction is performed on local point cloud data through the Pointnet++ grid structure, including multi-layer perception and feature extraction of local point clouds.

It should be noted that the Pointnet++ network is used for feature extraction. By stacking multiple Pointnet++ modules, the point cloud data can be projected onto a specific space (nodes of the octree) and extracted using multiple MLPs (Multilayer Perceptrons). Features, and use max pooling operations to aggregate local features, you can gradually extract multi-scale features of point cloud data, which will provide feature representation with rich semantic information for each point of the point cloud; you can use hierarchical aggregation operations to obtain a more global Contextual information, the network can make full use of local and global information for more accurate segmentation tasks.

S4: Use the fully connected layer to predict the category of each point. Based on the prediction results, segment and semantically label the point cloud, and conduct training and optimization, including,

For specific point cloud segmentation tasks, the segmentation head network is introduced in Pointnet++ to output the semantic segmentation label of each point, and the segmentation head is adjusted according to the task requirements;

In the training phase, point cloud data with known labels are used for supervised learning, and the parameters of the network are updated through backpropagation and optimization algorithms, with the goal of minimizing the loss function between the predicted labels and the true labels.

S5: Use the trained network to segment the new real-life point cloud and obtain the semantic label of each point to segment, extract and analyze different objects based on the label, including,

In the testing phase, the trained Pointnet++ network model is used to segment the new point cloud data, forward propagation is performed on each point, and the corresponding label is assigned to the point based on the final segmentation result.

It should be noted that directly processing Pointnet++ on the entire high-resolution point cloud may result in excessive computational complexity. Our solution Pointnet++ can operate more efficiently on each local area, reducing computing and memory requirements. At the same time, through the hierarchical structure, the level of segmentation operations can be selected in a targeted manner to further improve calculation efficiency.

The above is a schematic solution of an automatic object segmentation method based on a real point cloud of a substation in this embodiment. It should be noted that the technical solution of the system for automatic segmentation of objects based on the real-life point cloud of a substation belongs to the same concept as the above-mentioned technical solution of the automatic segmentation method of objects based on the real-life point cloud of a substation. In this embodiment, the system based on the real-life point cloud of a substation belongs to the same concept. For details that are not described in detail in the technical solution of the automatic object segmentation and push system, please refer to the above description of the technical solution of the automatic object segmentation method based on the real point cloud of the substation.

In this embodiment, the object automatic segmentation system based on the real point cloud of the substation includes:

The acquisition module is used to obtain real-life point cloud data and corresponding annotation information;

The preprocessing module is used to preprocess real-life point cloud data through an adaptive layering method. The preprocessing divides the point cloud data into different levels and arranges the organization of points;

The fusion module is used to extract features from point clouds at each level and fuse point cloud features at different levels to obtain global and comprehensive feature representation;

The training module is used to predict the category of each point using a fully connected layer, segment and semantically label the point cloud based on the prediction results, and perform training and optimization;

The segmentation module is used to use the trained network to segment new real-life point clouds and obtain the semantic labels of each point to segment, extract and analyze different objects based on the labels.

This embodiment also provides a computing device suitable for automatic segmentation of objects based on real-life point clouds of substations, including:

Memory and processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the automatic object segmentation method based on the real point cloud of the substation as proposed in the above embodiment.

This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by the processor, the automatic object segmentation method based on the real point cloud of the substation is implemented as proposed in the above embodiment.

The storage medium proposed in this embodiment and the method for automatic object segmentation based on real-life point clouds of substations proposed in the above embodiment belong to the same inventive concept. Technical details that are not described in detail in this embodiment can be found in the above embodiment, and this embodiment It has the same beneficial effects as the above embodiment.

Through the above description of the implementation, those skilled in the art can clearly understand that the present invention can be implemented with the help of software and necessary general hardware. Of course, it can also be implemented with hardware, but in many cases the former is a better implementation. . Based on this understanding, the technical solution of the present invention can be embodied in the form of a software product in essence or that contributes to the existing technology. The computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk. , read-only memory (ReadOnly, Memory, ROM), random access memory (Random Access Memory, RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions to make a computer device (can be a personal computer, server , or network equipment, etc.) to perform the methods of various embodiments of the present invention.

Example 2

Referring to Table 1 to Table 3, an embodiment of the present invention provides an automatic object segmentation method based on real point clouds of substations. In order to verify its beneficial effects, an exemplary simulation application scenario and comparison scheme are provided for scientific Argument.

Obtain real-life point cloud data and corresponding annotation information;

The real-life point cloud data is preprocessed through the adaptive layering method. The preprocessing divides the point cloud data into different levels and arranges the organization of points;

Perform feature extraction for each level of point cloud, and fuse point cloud features at different levels to obtain global and comprehensive feature representation;

Use a fully connected layer to predict the category of each point, segment and semantically label the point cloud based on the prediction results, and perform training and optimization;

Use the trained network to segment the new real-life point cloud and obtain the semantic label of each point to segment, extract and analyze different objects based on the label.

Based on the above scheme, the specific patterns in some substation hierarchies are shown, see Table 1.

Table 1 Specific modes in some substation hierarchies

In the testing phase, the trained Pointnet++ network model is used to segment the new point cloud data and assign corresponding labels to each point. Table 2 is an example showing the results of segmenting point cloud data using Pointnet++:

Table 2 Segmentation results

Table 2 shows the results after segmenting the sample point cloud data. According to the predicted label, the functional area or device type to which each point belongs can be determined.

Among them, S2 involves the selection of point cloud organization methods, which are compared through simulation data, see Table 3.

Table 3 Data examples of local neighborhood point cloud method in substation simulation scenario

In the above table, local density is the power demand density or personnel density in a certain area of the substation. Vibration amplitude is used to evaluate the vibration level of substation equipment. Current fluctuation is used to describe the instability of the grid load. Temperature fluctuations are used to evaluate temperature changes in substation equipment. The category label represents the object category in the point cloud, such as transformers and insulators.

By selecting the local neighborhood point cloud, you can see that for the same category (transformer or insulator), the parameter values of the local neighborhood point cloud are smaller. This shows that the local neighborhood point cloud method has a better ability to capture specific targets in substation simulation scenarios.

It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. 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 carried out. Modifications or equivalent substitutions without departing from the spirit and scope of the technical solution of the present invention shall be included in the scope of the claims of the present invention.

Claims (10)

1. An automatic object segmentation method based on real-life point clouds of substations, which is characterized by: Obtain real-life point cloud data and corresponding annotation information; The real-life point cloud data is preprocessed through an adaptive layering method. The preprocessing divides the point cloud data into different levels and arranges the organization of points; Perform feature extraction for each level of point cloud, and fuse point cloud features at different levels to obtain global and comprehensive feature representation; Use a fully connected layer to predict the category of each point, segment and semantically label the point cloud based on the prediction results, and perform training and optimization; Use the trained network to segment the new real-life point cloud and obtain the semantic label of each point to segment, extract and analyze different objects based on the label. 2. The object automatic segmentation method based on the real-life point cloud of a substation as claimed in claim 1, wherein the obtaining the real-life point cloud data and the corresponding annotation information, wherein the obtaining the real-life point cloud data includes: Obtain information on the geometry and location of buildings, equipment, power lines, and heat distribution; The corresponding annotation information includes category annotation, bounding box annotation, and point cloud annotation. 3. The object automatic segmentation method based on the real-life point cloud of a substation as claimed in claim 1 or 2, characterized in that the real-life point cloud data is preprocessed through an adaptive layering method, and the preprocessing converts the point cloud into The way data is organized into different levels and arranged points, including, Through the octree method, a hierarchical structure is constructed according to the distribution and characteristics of the point cloud to obtain point clouds of different levels, including setting four levels; The first level includes cloud data of the entire substation, providing global characteristics of the substation; The second level divides the substation into different heat distribution areas based on the point cloud data of heat distribution; The third level divides substation site cloud data into different equipment functional areas based on the distribution and complexity; The fourth level divides the grid according to the type and function combination of the equipment inside each area. 4. The object automatic segmentation method based on the real point cloud of the substation as claimed in claim 3, characterized in that the arrangement of points is used to obtain the point clouds of different levels with different densities and resolutions. Specifically, include, In the first level and the second level, the first neighborhood radius threshold is set through the local neighborhood point cloud method to determine the corresponding local neighborhood; In the third level and the fourth level, the second neighborhood radius threshold is set through the local neighborhood point cloud method to determine the corresponding local neighborhood; The first neighborhood radius threshold is greater than the second neighborhood radius threshold. 5. The object automatic segmentation method based on the real point cloud of the substation as claimed in claim 4, further comprising, after determining the local neighborhood, setting the resolution according to the relative position of the point or projection to different grids. The resolution of each level; The first level and the second level set the first resolution; The third and fourth levels set the second resolution; The first resolution is smaller than the second resolution; In the hierarchical structure, for each level of point cloud data, the local coordinates relative to that level are calculated for use in the segmentation stage. 6. The object automatic segmentation method based on real point clouds of substations as claimed in claim 5, characterized in that feature extraction is performed on point clouds at each level, and point cloud features at different levels are fused to obtain global and comprehensive features. The characteristic representation includes, At each level after adaptive layering, feature extraction is performed on local point cloud data through the Pointnet++ grid structure, including multi-layer perception and feature extraction of local point clouds; Use a fully connected layer to predict the category of each point. Based on the prediction results, the point cloud is segmented and semantically labeled, and trained and optimized, including, For specific point cloud segmentation tasks, the segmentation head network is introduced in Pointnet++ to output the semantic segmentation label of each point, and the segmentation head is adjusted according to the task requirements; In the training phase, point cloud data with known labels are used for supervised learning, and the parameters of the network are updated through backpropagation and optimization algorithms, with the goal of minimizing the loss function between the predicted labels and the true labels. 7. The object automatic segmentation method based on the real-life point cloud of the substation as claimed in claim 1 or 6, characterized in that the trained network is used to segment the new real-life point cloud to obtain the semantic label of each point, so as to obtain the semantic label of each point based on the trained network. Tags segment, extract and analyze different objects, including, In the testing phase, the trained Pointnet++ network model is used to segment the new point cloud data, forward propagation is performed on each point, and the corresponding label is assigned to the point based on the final segmentation result. 8. A system for automatic segmentation of objects based on real-life point clouds of substations, characterized by: The acquisition module is used to obtain real-life point cloud data and corresponding annotation information; A preprocessing module for preprocessing the real-life point cloud data through an adaptive layering method. The preprocessing divides the point cloud data into different levels and arranges the organization of points; The fusion module is used to extract features from point clouds at each level and fuse point cloud features at different levels to obtain global and comprehensive feature representation; The training module is used to predict the category of each point using a fully connected layer, segment and semantically label the point cloud based on the prediction results, and perform training and optimization; The segmentation module is used to use the trained network to segment new real-life point clouds and obtain the semantic labels of each point to segment, extract and analyze different objects based on the labels. 9. An electronic device, including: memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the real-time point cloud based on the substation in any one of claims 1 to 7 is realized. The steps of automatic object segmentation method. 10. A computer-readable storage medium that stores computer-executable instructions. When executed by a processor, the computer-executable instructions implement the object automatic segmentation method based on real-life point clouds of substations described in any one of claims 1 to 7. step.