Comparative analysis of relative motion estimation errors under varying motion scenarios

Accurate motion estimation is a fundamental task in computer vision, robotics, and autonomous systems. It is critical for applications such as autonomous navigation, vehicle localization, and pose tracking in augmented and virtual reality environments. Although numerous linear and nonlinear algorithms have been developed, their performance under different motion conditions remains insufficiently studied. We present a comparative evaluation of four classical motion estimation models: essential matrix decomposition, homography matrix decomposition, photogrammetric relative orientation, and homography-based relative orientation. The models are tested on real and synthetic monocular image sequences of a rigid object undergoing controlled motion, including pure rotation, translation-dominated motion, and combined motion. All models are evaluated using a unified preprocessing pipeline to ensure consistency. The results indicate that homography-based models are more robust under planar motion, whereas the essential matrix and relative orientation methods yield better accuracy in depth-varying translational scenarios. These findings offer practical guidance for selecting appropriate motion estimation techniques in vision systems where robustness and computational efficiency are essential.