| Li DAR(light detection and ranging)can directly and quickly obtain high-density and high-precision three-dimensional point cloud data,having extensive application in survey,digital city,power inspection and other fields.With the improvement of Li DAR system performance,acquisition and process of massive point cloud have become more and more common,challenging the computer processing ability and network transmission ability.Therefore,the reasonable data organization,compression and visualization has become the key to the efficient application of point cloud on the web.However,unlike regular grid data such as remote sensing images,point cloud data is strong discreteness and unstructured,which has higher requirements for data organization index,compression transmission and visualization algorithm.This paper carries out the research on the organization,compression and visualization of massive point cloud data for the Web.This paper takes the application of Li DAR power line inspection as an example,building a Li DAR data visualized management and analysis platform for power line inspection.The main contents are as follows:(1)A visualization oriented progressive point cloud organization and lossless compression algorithm is proposed.This method realizes the rapid index of massive point cloud through chunk parallel octree index and improved adaptive random sampling.Combined with spatial filling curve,differential coding and adaptive arithmetic coding,it realizes point cloud data compression and supports levels of detail based visualization.The experimental results show that the index algorithm can generate more uniform samples,and the efficiency and compression ratio are improved by 44% and 71% respectively compared with that of LAZ algorithm.(2)A point cloud compression algorithm based on deep variational auto-encoder is proposed,which realizes multi-scale feature fusion combined with attention mechanism to establish a multi-scale deep entropy model and combines entropy coding to realize point cloud lossy compression with high bit rate and low distortion,and solves the problem of voxel class imbalance in the training process by focal loss.Experiments and analysis are carried out based on MPEG data sets.The results show that compared with traditional G-PCC and deep-learning based D-PCC algorithms,D1 BD-PSNR is improved by 2.81 d B and 4.67 d B respectively,and D2 BD-PSNR is improved by2.66 d B and 4.58 d B respectively.(3)Research on point cloud rendering and visual management platform on the Web.Web GL,Levels of Detail and frustum culling technology are used to realize real-time loading and efficient rendering of point cloud on the Web.Based on the point cloud index and compression visualization algorithm,this study builds a visual management and analysis platform for Li DAR power line inspection on the Web,which can provide point cloud browsing and interactive management services for power line,and have the functions of point cloud data storage,compression management,efficient visualization,power line features management and interactive analysis. |
