A novel quality control procedure for the evaluation of laser scanning data segmentation
Over the past few years, laser scanning systems have been acknowledged as the leading tools for the collection of high density 3D point cloud over physical surfaces for many different applications. However, no interpretation and scene classification is performed during the acquisition of these datasets. Consequently, the collected data must be processed to extract the required information. The segmentation procedure is usually considered as the fundamental step in information extraction from laser scanning data. So far, various approaches have been developed for the segmentation of 3D laser scanning data. However, none of them is exempted from possible anomalies due to disregarding the internal characteristics of laser scanning data, improper selection of the segmentation thresholds, or other problems during the segmentation procedure. Therefore, quality control procedures are required to evaluate the segmentation outcome and report the frequency of instances of expected problems. A few quality control techniques have been proposed for the evaluation of laser scanning segmentation. These approaches usually require reference data and user intervention for the assessment of segmentation results. In order to resolve these problems, a new quality control procedure is introduced in this paper. This procedure makes hypotheses regarding potential problems that might take place in the segmentation process, detects instances of such problems, quantifies the frequency of these problems, and suggests possible actions to remedy them. The feasibility of the proposed approach is verified through quantitative evaluation of planar and linear/cylindrical segmentation outcome from two recently-developed parameter-domain and spatial-domain segmentation techniques.