Ploughing furrow recognition for onland ploughing using a 3D-LiDAR sensor

Abstract

The purpose of the work presented in this paper is to develop a robust method of autonomous vehicle localisation during soil tilling, using a 3D LiDAR sensor. The approach was developed for onland ploughing but is transferable to other applications with similar features. To detect a ploughing furrow from multiple layers in a 3D point cloud, a recognition method was employed consisting of two parallel and redundant branches: (1) detecting the furrow edge in each layer by identifying the characteristic step, and (2) approximating the unploughed surface by applying Euclidean clustering and linear least-squares line fitting. The redundant branches were then fused into a guidance directrix by RANSAC line fitting. To evaluate the algorithm, human-labelled reference points of the furrow were obtained by elaborate manual labelling of the LiDAR ploughing data. The results of the recognition method were compared with the reference points, resulting in a minimum of 86 % identical recognised points, with a mean error of 6.5 mm.