실내에서의 LiDAR와 이미지의 라인 정보를 활용한 위치 보정

Abstract

Since absolute position sensors such as GPS cannot be used in the indoor environment, it is necessary to use existing information to know the absolute location. If there aren't any pre-built maps, the only way to get this information is from the floor plan. In this paper, we propose a method to use the 3D line information extracted from the LiDAR sensor and the line information extracted from the image data to correct the current position by using the line information existing in the floor plan. Absolute matching with the map is possible using the 3D lines extracted by normal difference calculation and RANSAC, and the orientation of the robot can be corrected by comparing the vanishing point calculated from the line of the floor plan projected on the image with the line information extracted from the image. Through this algorithm, the absolute position can be robustly compensated even with the lack of point features or existence of dynamic obstacles in the environment.