大規模点群データの平滑化手法に関する研究 (第1報)

—ロバスト推定に基づく平滑化手法—

抄録

It is widely recognized that task planning based on 3D CAD can reduce the rework of maintenance and renovation of facilities. Therefore, it is very important to acquire 3D shapes of existing facilities. The state-of-the-art phase-based scanner is suitable for this purpose, because it can acquire hundreds of millions point data in several minutes. However, point data captured from the phase-based scanner tend to include quite a lot of outliers. This paper introduces robust smoothing operators for noisy point-cloud. We propose two smoothing methods using robust estimate. One is based on Lorentzian distribution, and the other is based on Tukey's bi-weight estimation. We modified a conventional smoothing method using robust estimate. In our experiments, our two methods could produce good smoothed surface even if point-cloud include a lot of outliers.