In this paper we present an efficient edge detection algorithm for the extraction of linear features in both range and intensity image data. The purpose is to simplify the dense datasets and to provide stable features of interest, which are used to recover the positions of the 2D cameras with respect to the geometric model for tasks such as texture mapping. In our algorithm the required features of interest are extracted by an analysis of the mean curvature values. As it will be demonstrated, the algorithm features computational efficiency, high accuracy in the localization of the edge points, easy implementation, and robustness against noise. The generality and robustness of the algorithm is llustrated during processing of complex cultural heritage scenes.
Artikelauszug / Extract:
1. Introduction
Terrestrial laser scanning has become a standard tool for 3D data collection to generate high quality 3D models of cultural heritage sites and historical buildings [Boehler and Marbs, 2002]. Based on the run-time of reflected light pulses these systems allow for the fast and reliable measurement of millions of 3D points allowing for a very effective and dense measurement of the surface geometry. In addition to the geometric data collection, texture mapping is particular important in the area of cultural heritage in order to allow a complete documentation of the respective sites. For this reason, some commercial 3D systems provide model-registered color texture by the simultaneous collection of RGB values of each LIDAR point. For this purpose a camera is directly integrated in the system. However, the ideal conditions for taking images may not coincide with those for laser scanning [El-Hakim et al, 2002].