A Simple Visual Odometry Approach Based on Point Clouds

In this paper, a simple approach for estimating the ego-motion of a vehicle is proposed. The system can estimate the motion based on a sequence of images and point cloud data. The system is capable of estimate the ego-motion through a 2-D to 3-D mapping of the detected features on the image considering that the calibration parameters of the sensors are available. Singular value decomposition is used to estimate the rigid-body transformation between two point clouds, and we employ a RANSAC outlier rejection method to estimate a more accurate ego-motion. Datasets from the KITTI Benchmark Suite are used in order to evaluate the proposed method. The experiments show that our approach performed well with translational errors below 5%.