Robust and Diagnostic Statistics: A Few Basic Concepts in Mobile Mapping Point Cloud Data Analysis

Abstract

It is impractical to imagine point cloud data obtained from laser scanner based mobile mapping systems without outliers. The presence of outliers affects the most often used classical statistical techniques used in laser scanning point cloud data analysis and hence the consequent results of point cloud processing are inaccurate and non-robust. Therefore, it is necessary to use robust and/or diagnostic statistical methods for reliable estimates, modelling, fitting and feature extraction. In spite of the limitations of classical statistical methods, an extensive literature search shows not much use of robust techniques in point cloud data analysis. This paper presents the basic ideas on mobile mapping technology and point cloud data, investigates outlier problems and presents some applicable robust and diagnostic statistical approaches. Importance and performance of robust and diagnostic techniques are shown for planar surface fitting and surface segmentation by using several mobile mapping real point cloud data examples.