Evaluate 3D laser point clouds registration for cultural heritage documentation

According to the International Council on Monument and Sites (ICOMOS), cultural heritage is defined as an expression of the manners of living improved by a community and passed on from past generation to future generation. Three-dimensional point cloud registration is one of the critical topics in close-range remote sensing and particularly in laser scanning. Laser scanning systems in most situations are restricted to scan a segment of the interested site; nevertheless in the greatest majority of applications a scanning for the entire site is required. Iterative closest point (ICP) is an important algorithm to achieve fine registration for two or more overlapping range data by iteratively approximating the Euclidean distances. Many authors, in recent years had been proposed different approaches to improve and modify the ICP registration technique for optimum outputs. In this paper a comparison study was presented to assess Nearest Neighbor Iterative Closest Point (NN-ICP) and Levenberg-Marquardt Iterative Closest Point (LM-ICP) registration methods in order to evaluate the 3D close-range TLS data outputs for cultural heritage documentation in IRAQ. Additionally, the study is assessing the accuracy of culture heritage documentation based on close-range TL scanning alongside processing elapse time as a priority factor. To investigate these objectives, two experiments were carried out on two study sites in Baghdad/IRQ. The average registration errors delivered from both case studies following automatic process were (3.9 mm) lasts for (49 s) and (2.6 mm) lasts for (200 s) for NN-ICP and LM-ICP respectively. In order to evaluate the accuracy of the extracted 3D digital models, conventional survey measurements were executed for a set of evenly distributed reference points mounted around the body of two selected objects in both sites. The RMSE delivered from the indoor and the outdoor sites were (6mm) and (12 mm) with minimum range of (3.5 m) and (7 m) respectively.