Improved real-time odometry estimation method for incremental RGB-D mapping by fusing IMU data

摘要

RGB-D cameras can provide rich 2D visual and 3D point cloud information simultaneously, which are well suited to estimate ego-motion. This paper proposes a robust and accurate visual odometry method by fusing inertial measurement unit(IMU) sensor data: we use GPU-based SIFT features for online odometry estimation and improve its stability by fusing IMU data. The pose of each RGB-D frame is applied to incremental mapping approach, and we realize the registration for over-lapped models by combining the features of key-frames and Simplified-ICP algorithm. Experimental results show the effectiveness and scalability of our proposed approach.