Line Feature Extraction from RGB Laser Point Cloud

Abstract

Line feature extraction from point cloud is a useful technology in many application fields such as surveying, 3d reconstruction and self-driving. Currently, existing methods focus on line feature extraction solely from point cloud data while less focus has been put into point cloud with RGB texture information. This paper proposes a line extraction method from RGB laser point cloud under the RANSAC framework. The extracted line features include the intersection lines between planes, line features with depth discontinuity and those with change in RGB intensity values. The developed algorithm adopted plane segmentation of point cloud, bit map construction, line segment detection with global RANSAC. The experimental results show that the majority of the line features can be extracted while being robust to point cloud noise, outliers and missing data.