두 점군의 초기 정합에서 GPS와 IMU 센서들의 활용 방법에 대한 연구

Abstract

This paper presents a method to utilize the GPS (Global Positioning System) and IMU (Inertial Measurement Unit) sensors for the initial registration of 3D scanned point clouds. First, two sets of point-clouds of a target scanned at two different positions and orientations are generated, and the scanner’s positions and attitudes corresponding to each point cloud are also obtained from the GPS and IMU sensors. Second, the data are post-processed by a Kalman filter and the initial registration is performed using the GPS and IMU sensor data. Third, fine registration of the two point clouds is carried out using the results from the initial registration using an ICP (Iterative Closest Point) algorithm. As a result, the number of iterations in the fine registration step is decreased by 88.8 percent, the processing time by 92 percent, and the root-mean-square error (RMSE) by 2.4 percent respectively, when the initial registration using the GPS data is applied. Likewise, the similar tests are performed to obtain more decreases by 19.9 percent for the number of iterations, 17.4 percent of processing time, and 2.9 percent of RMSE, respectively, when the GPS and IMU are combined for the initial registration.