Authors: Kaleel Al-Durgham, Ayman Habib, Mehdi Mazaheri
Over the last decade, terrestrial laser scanner systems have been proven to be an effective tool for the acquisition of 3D spatial information over physical surfaces. Many factors such as the low cost and the ability of rapidly collecting dense and accurate spatial data led to the utilization of laser scanners in different applications such as industrial sites modeling, 3D documentation of buildings, and many civilian and military needs. Usually, a complete 3D model for a given site cannot be derived from a single scan. Therefore, several scans with significant overlap are needed to cover the entire site and also to attain better information about the site than what could be obtained from a single scan. However, the collected scans will be referenced to different local frames that are associated with the individual scanner locations. Hence, a registration process, which aims at estimating the 3D-Helmert transformation parameters, should be established to realign the different scans to a common reference frame.
This paper introduces a new methodology for the automatic registration of terrestrial laser scans using linear features. Linear, cylindrical, and pole-like features are directly extracted from the scans through a region-growing procedure. Hypothesized conjugates of linear features are identified using invariant separation characteristics such as spatial separation and angular deviation between two linear features. All the hypothesized conjugate pairs – taken one at a time – are used to estimate the 3D-Helmert transformation parameters that are required to realign one scan to the reference coordinate system of another scan. Logically, only the right conjugate pairs among the hypothesized matches will lead to similar solutions of the transformation parameters. Therefore, we developed a strategy to detect the most frequent set of estimated parameters. A linear mathematical model that utilizes quaternions to represent rotation angles is used to simplify the estimation of the transformation parameters. Experiments will assess the performance of the proposed methodology over multiple scans of a power plant.
You are reviewing the
Download.
Your machine is locate at 3.142.133.41.