A new method for calibration of range sensor and terrain classification

Laser range scanner is an important modality in robotics for the perception of an environment such as object classification, map generation, navigation, etc. In this paper, we first present a new method for calibration of Laser range scanner to generate the highly accurate 3D model of the environment. Later, we propose a robust segmentation method for objects and ground points separation of fused range data which is obtained from a Laser range scanner. The Laser range scanner gives the two range dataset from different orientation of the same terrain. After fusion of these two range data using ICP algorithm, the complete range data of terrain are obtained. In order to classify objects and ground points of terrain, we exploit the properties of statistical measures and local elevation of the range data. The proposed algorithm is independent from range data format and resolution, i.e., it works for point cloud and gridded data. The experimental results presented in the paper have shown robustness of the proposed method for explicit segmentation of objects from the ground points.