Distortion Refining of Laser Point Cloud from In-vehicle Laser Scanner based on NDT Scan Matching

Abstract

This paper presents a Kalman-filter based method of refining distortion of 3D laser-scanned data from in-vehicle multilayer laser scanner. The vehicle identifies its own 3D pose (position and attitude) based on Normal-distributions-transform (NDT) scan-matching method. Based on the pose information, the vehicle pose in which the laser scan data is captured is predicted and smoothed using Kalman filter under the assumption that the vehicle moves at almost constant linear and turning velocities. The predicted and smoothed poses of the vehicle are applied to mapping laser-scanned data onto the world coordinate frame. The vehicle again identifies its own 3D pose based on NDT scan matching. Such an iteration enables to compensate the distortion of laser-scanned date. Experimental results show the performance of the proposed a method.