A Point and Line Features Based Method for Disturbed Surface Motion Estimation

Abstract

Calculating the motion of disturbed surface such as a reflective monochromatic one is often a difficult part, especially when using single feature based method. The error introduced from the feature extraction and matching will gradually accumulate into a larger final error. For a texture-less surface, the number of features makes the situation even more challenging. In this paper, point and line features from stereo sequences are combined to estimate 3D motion of disturbed surfaces. Taking the advantage of feature combination by two-stage iterative optimization and multiple filtering, the motion of surfaces can be estimated accurately, even under little motion blur. This paper also explored the relationship between measurement accuracy and object motion mode. This may provide a reference for the design of a vision based motion measuring system.