2018 年 84 巻 8 号 p. 738-745
3D laser scanners have recently been introduced for efficient as-built modeling in the heating, ventilating and air conditioning (HVAC) industry. Given their highly complex installation, when scanning the piping objects in HVAC systems using a terrestrial laser scanner (TLS), it is difficult to manually determine feasible scanner placement to capture the object areas required for renovation work at high accuracy and quality without occlusion. We propose an algorithm for a model-based optimum scan planning method for a TLS. This method uses a coarse 3D model generated by structure-from-motion, and finds optimal scanner placement that satisfies visibility, scan range, incident angle, and scan quality constraints. In this study, the optimal scanner placement problem is formulated as a 0-1 integer programming problem that is solved using the branch and bound method. In addition, a GPU significantly increases the speed of the precomputation of visibility constraints. In an evaluation, the proposed algorithm outperformed our previously proposed greedy method in terms of the number of scans and scan coverage ratio, and it shows performance equal to the greedy method in terms of modeling quality.