Electrical Equipment Classification Based On 3D Point Cloud

With the rapid development of power industry, more requirements are put forward for the substation construction, and the digital reconstruction of substation is an important part of the monitoring and diagnosis. The digital reconstruction model of substation based on the outline and the posture information of substation equipment can reflect the real situation of the substation equipment space position information, and it is convenient for maintenance and upgrading of substation and the layout adjustment, at the same time, it can provide more accurate basis for the substation reconstruction.Classification and identification of electrical equipment is the key technology of digital reconstruction in substation, which belongs to 3D laser point cloud. We could scan the substation equipment and extract its features from 3D point cloud for substation equipment species classification, which is quite necessary for automatic digital reconstruction.In this thesis, we investigate the issues of feature extraction and substation equipment species classification as well as pose estimation problem, the main research contents are as follows.(1) This thesis introduces the working principle of 3D laser scanning system and the acquisition and pretreatment process of the point cloud of substation equipment.(2) We study on the extraction principle and method of substation equipment based on 3D point cloud data, and the feature of layered projection area is proposed,then we extract the vector characterize of substation equipment based on the principal component analysis, at the same time, projection area, equipment projection density, gray image, envelope volume are extracted, which lays a good foundation for the classification and identification of equipment.(3) Compare the support vector machine 、 neural network and distance classification algorithm, we choose the distance classification algorithm as our classification algorithm, because of high feature dimension. we put forward an improved method of PSO, then use the algorithm to optimize feature weights, which improve the classification accuracy of the equipment.(4) Select volume matching algorithm to estimate the posture of the substation equipment, which provides the position information for the automatic 3D digital reconstruction of the substation. At the same time, we establish a simulation experiment platform in the MATLAB environment, which can display the process of classification of substation equipment intuitive.The classification and recognition algorithm of substation equipment based on3 D point cloud is studied in this thesis, and the point cloud data analysis and research are carried out on 12 kinds of 222 kinds of substation equipment, then the recognition feature of the device is extracted, and the simulation test is carried out.The results show that this algorithm can realize automatic recognition of substation equipment and attitude estimation, which laid the foundation for the automatic reconstruction of digital substation. At the same time, the content of this thesis can improve the efficiency of 3D digital substation reconstruction, and improve the efficiency.