2D feature extraction in sensor coordinates for laser range finder

Abstract

This paper describes an algorithm for feature extraction from measurements collected by laser range finder. The features are straight lines and corners present in the environment. The measurements are processed in the sensor coordinate system to identify an invariant parameter for the present straight lines that indicate their position and orientation. A second parameter is needed to determine the slopes of the straight lines as well as the values of the intersection angles between the consecutive lines which can be considered as corners in the case of right angles. A log-Hough transformation is used to find the equations of the straight lines. All the computations are carried out in the sensor coordinate system which leads to a fast and robust algorithm of feature extraction. Practical examples are presented to evaluate the efficiency of the algorithm.