How to Find Geometric Changes in Laser Scanning Point Clouds? A Perspective on Correspondence Definitions

Abstract

3D point clouds generated from laser scanning techniques offer opportunities for precise and efficient reality capture with higher spatial resolution compared to traditional point-wise techniques. The consequent 3D change detection and analysis based on multitemporal point clouds have seen rapid advancements over the past two decades. In this context, numerous methods have been proposed to detect and analyze surface changes in general or specific scenarios. This paper systematically reviews and illustrates various methodologies for change analysis based on laser scanning point clouds, focusing particularly on the definitions of correspondences. These correspondences between compared point clouds are defined according to the types of changes that are expected to be detected, including surface differences, displacement vectors, and parametric changes, which result in different analytical approaches. Using bitemporal laser scanning point clouds of a rock glacier surface, we demonstrate and evaluate the impact of different methods on quantified changes and provide suggestions for selecting appropriate methods across different application scenarios. Additionally, we highlight existing challenges and research directions for advancing change analysis using laser scanning point clouds.