Study On Fast 3D Modeling Method Of Substation Based On Point Cloud Data

The substation is a key part to maintain the normal operation of the power system.The three-dimensional real scene simulation model of the substation with real information,excellent model pictures,and high scalability.And the model is the realistic basis for constructing digital and intelligent power systems,virtual reality simulation training systems,and augmented reality on-site auxiliary analysis systems.The scene model built with the point cloud data collected by the three-dimensional laser scanning equipment has high realism and spatial accuracy.Because the point cloud data of the substation has the characteristics of large data volume,uneven distribution,excessive background noise,etc.,the traditional point cloud data processing method applied to the substation modeling has problems such as residual noise,difficult equipment segmentation,and low model matching efficiency.Therefore,this paper deeply researches the key technology based on point cloud data modeling,and proposes an automatic modeling method,which solves the difficulties of cloud data processing at substation and improves the modeling speed.The main research work and results of this paper are as follows:To solve the problem of large data volume and uneven distribution of point cloud data sets of substation scenes,this paper proposes an adaptive voxel filtering algorithm to calculate the neighborhood density of each point.In this method,the density is used as the criterion for judgment.The point cloud in the density area is processed with different compression parameters to ensure a high compression ratio while retaining low-density details inside the device.In view of the large amount of high-density background noise inside the substation scene,there is still a problem of residual noise after denoising.This paper proposes a denoising algorithm based on longitudinal constrained hierarchical clustering.By introducing height features,it is based on the clustering algorithm.Comprehensive consideration of the horizontal and vertical distribution of the point cloud data to effectively distinguish the point cloud data of the electrical equipment from the point cloud data of the background noise,and effectively delete more than 97% of the highdensity noise.Aiming at the difficulty of segmenting the point cloud data of electrical equipment in substations,this paper proposes an electrical equipment extraction method based on neighborhood feature aggregation,designs a neural network model based on the codec structure,and uses coding and decoding modules to extract and strengthen the substation Point cloud data neighborhood spatial features,feature learning module learns neighborhood features,realizing fast and accurate segmentation and extraction of electrical equipment,segmentation accuracy is as high as 97%.For improving the efficiency of the model matching,a fast matching method based on image recognition is proposed.First,build a model based on convolutional neural network,perform feature extraction and feature learning of electrical equipment images,and preliminarily classify and recognize individual electrical equipment point cloud data after segmentation.The recognition accuracy rate is as high as 99%.Then,the classified electrical equipment point cloud model is matched with the standard model library model,the corresponding position of the standard model and the real equipment in the scene is obtained,and the scene model is reconstructed.This method can reduce the amount of matching calculation and greatly improve the reconstruction speed.Finally,using the point cloud data sets of the three voltage levels of Maohu Substation as the experimental object,the experiment verifies that the modeling method proposed in this paper can realize the rapid modeling of the three-dimensional real scene simulation substation.