On segmentation, shape estimation and shape estimation and navigation using 3D laser range measurement of forest scenes

摘要

The problem addressed is segmentation and shape estimation of three-dimensional laser range measurements of forest scenes. For each tree the position as well as the three dimensional shape of the trunk is estimated. The detection uses a two-dimensional histogram of the measurements, followed by segmentation by fitting cylinders around the measurements of each detected tree trunk. The shape estimation consists of fitting a 3D spline to each tree trunk's centre curve and computing the trunk width by extracting the range discontinuities at the edges. For the experiments a mobile robot carrying a scanning laser range finder was used. Navigation was done by matching the extracted tree trunk positions from subsequent range scans, determining the relative movements.