An improvement method of wall points extraction process from mobile mapping system data by localization method of space division

摘要

A purpose of our research is to generate three-Dimensional (3D) building models since they are always the most important elements for landscape simulation and urban planning. The authors have proposed an extraction method for individual building wall points from a Mobile Mapping System data as the first step for generating the virtual space. However, this wall points extraction method has some problems. For example, it might extract undesirable planes which are different from walls of buildings and it is also time-consuming for extracting small planes by using Random Sample Consensus (RANSAC). This paper gives the reason why the RANSAC in our system extracts undesirable planes under influence of the structure which are far away from it, and also show the reason that the inlier ratio becomes lower as the plane extraction process goes on. To solve these problems, the authors proposed the idea of localizing the extracting target from the influence factors which are far away from the target. It shows that this method has highly improved the extraction results as well as largely reduced the processing time.