Normal Vector Direction-based 3D LiDAR Point Cloud Planar Surface Removal for Object Cluster Minimization in Human Activity Monitoring System

The use of 3-Dimensional Light Detection and Ranging (3D LiDAR) point cloud as the alternative data to reduce privacy exposure in monitoring systems has been carried out in several studies. Unfortunately, various challenges in using point clouds intersect with the amount of data and computational costs. Several studies attempted to optimize the point cloud processing approach by segmenting the ground plane to get the object clusters separated. However, many unnecessary points can still burden the computation process. Since the ground plane mainly represents the horizontal planar plane on the x, y axis, this study tried to reduce the points on the vertical planar plane on the x, z and y, z axes with the x, y horizontal planar plane as well based on the surface normal vector direction of each point. The proposed approach has successfully reduced the raw point cloud by 79.29% removing the point cloud that indicates the planar surface of the three axes while maintaining the essential object of the monitoring system on the KITTI raw dataset. Therefore, the object cluster can be minimized, supporting the computational costs for further research in human activity monitoring systems.