Managing Huge Point Cloud Data through Geometrical-Based Registration

Point cloud data generated by laser scanners often comes in a very large file size, as it carries highly dense information of the scanned area. One of the ways to control and manage this huge data size is during the data collection process, where it can be conducted in smaller area of scanning, which can reduce the file size, and this is appreciated especially when dealing with bigger scale projects like building or life size monuments scanning. However, adapting this approach would require suitable registration methods in order to integrate and match the data together before further processing and reconstruction. This paper highlights on an automatic method of geometrical-based point cloud data registration, which work best when the scanning area are collected in half. In here, the geometrical entity representing both point clouds that need to be registered (i.e. straight surface) will be detected, and angle between the differences of these surfaces will be calculated before the second point cloud can be adjusted accordingly to be matched and registered with the first point cloud data. Experimental results on real point cloud data representing building interior shows that the algorithm works well in registering point cloud data together, hence making the process of handling huge point cloud data to be more manageable.