Advanced Computing: An International Journal

Recent decades have witnessed a significant increase in the use of visual odometry(VO) computer vision area. It has also been used varieties robotic applications, for example on Mars Exploration Rovers. This paper, firstly, discusses two popular existing odometry approaches, namely LSD-SLAM and ORB-SLAM2 to improve performance metrics SLAM systems using Umeyama Method. We carefully evaluate methods referred above three different well-known KITTI datasets, EuRoC MAV dataset, TUM RGB-D dataset obtain best results graphically compare evaluation from approaches. Secondly, we propose an approach running real-time with stereo camera, which combines feature-based (indirect) method feature-less (direct) matching accurate semidense direct image alignment reconstructing 3D environment directly pixels that gradient. Keywords VO, metrics, Method, method, & semi-dense real-time.