Research Of Algorithm Of Point Cloud Reconstruction Based On3-dimension Laser Scanning

Recently, with the development in laser scanning techniques, point cloud is widely applied to metrology and industrial control fields. Because of the flexibility in data storage and data processing, point cloud arises as an important metadata in computer graphics area.Reconstruction is an important technique in point cloud computing. With the development of 3D scanner and the scanned object surfaces being more complex, massive point data emerges, which brings new challenges for point cloud reconstruction.Based on the deep study of the technologies about point cloud reconstruction, this research mainly focus on the key technologies and problems in massive point cloud data reconstruction based on 3 dimensional laser scanner. Firstly, the 3D scanner is used for data acquisition, and the measure object information is stored as point cloud data. Based on the point cloud data, this paper developed uncertainty analysis and data pre-processing,including data segmentation and reduction, and then recovered the information of the measure object. The main contents are as follows:(1) The research of the uncertainty representation of point cloud. This paper focus on the relationship of the inaccurate data, deviations of scanner and bad measure conditions in the actual measurement process. To quantitatively measure the inaccuracy of the data,this paper discusses an uncertainty analysis model. Based on the Bayesian theory, the uncertainty of data can be quantitatively measured.(2) The research of data segmentation algorithm based on point cloud. Because of the huge data processing in K-means, the main restriction exists in time consumption during the cluster algorithm performing. Facing a large volume of point data, this paper discusses the data segmentation technique using K-means cluster algorithm. In order to reduce the time consumption, and reduce the number of iterations, this paper introduces a density bound to K-means algorithm. Based on K-means cluster, the estimation of cluster density and the adjust of center of cluster have been discussed.(3) The research of point cloud reduction algorithm for reserved features. In order to compress data volume, and improve the efficiency of surface reconstruction, this paperdiscusses the point cloud reduction algorithm. Starting with local differential geometric feature, this paper introduces natural quadratic surface as reduced model. Then, by analyzing the relationships of the model and point cloud, based on the feature of model surface, the point cloud reduction algorithm performed hierarchically has been discussed.(4) The research of the surface reconstruction algorithm. In order to get precise surface of measure object, this paper discusses the Poisson surface reconstruction algorithm. The efficiency of the Poisson algorithm and the quality of the surface has been concerned with the depth of octree. Besides, the bad result exists on the holes of surface.Point to these problems, based on data segmentation and the feature of surface, this paper discusses the new Poisson algorithm based on holes-repair using the greedy triangularization.At last, building the experimental platform by cantilever crane and 3D laser scanner,this research studied the implementation method of 3D point cloud reverse reconstruction system for reconstructing material piles. Using the collected data of material piles, this study verified the data acquisition and data pre-processing, and the realized the reconstruction process.