Research On Point Cloud Data Processing Technology In Grain Storage Measurements Using Three-Dimensional Laser Scanning

Laser scanning or LiDAR is an active remote sensing technology which has the advantage of directly measuring the three-dimensional (3D) coordinates of the points from a large-scale scene. It can be applied in many fields like machine vision, landscape scanning,3D reconstruction, etc. In this thesis, this technology is utilized to perform 3D modeling and volume measurement of bulk grain stored in large bins.The innovations and contributions of this thesis are mainly reflected in the following aspects: 1. This thesis proposes an efficient point cloud registration scheme suitable forgranary measurement applications. A feature extraction algorithm based on SIFT (Scale-Invariant Feature Transformation) technology with FPFH (Fast Point Feature Histograms) descriptors is adapted for the 3D point cloud scenario and used for coarse registration. The fine registration stage employs the famous ICP (Iterative Closest Point) algorithm. In order to make the conventional SIFT technology applicable to point cloud data containing only the 3D coordinates information, a modified SIFT algorithm adopting the point’s surface curvature as an additional feature filed is presented to achieve the task of feature search. 2. This thesis presents a new visualization design based on OpenGL (OpenGraphics Library) to render, display and manipulate 3D point cloud data. Comparing to the visualization APIs provided by Point Cloud Library (PCL), this design can support more formats of point cloud data and can be easily integrated with other application software. It can also provide extra rendering functions such as height-based rendering and class-based rendering. Experiments show that this design performs better than that using PCL visualization APIs. 3. In this thesis, a set of algorithms covering the entire flow for point cloud dataprocessing is designed and an interactive computer software system for 3D modeling and volume measurement of grain storage is also developed. The software is tested and validated using field data collected from the Qingyuan National Grain Storehouse located in Hebei Province of North China. Test results show the software can perform 3D modeling and volume calculation quickly and precisely. Comparing the data generated by the volume calculation module of the software system to those obtained by means of artificial measurement, the errors are less than 2%. In conclusion, the methods and software system presented in this thesis can work effectively and has good potential for use in the area of grain inventory.