抄録
This paper presents an algorithm that tracks rigid body motion using SIFT (Scale-Invariant Feature Transform) and voting technique. This algorithm clips target's area corresponding to feature points between two images. First, SIFT detects corresponding feature points between two images. Next, the algorithm determines a representative point for ROI (region of interest) by voting corresponding feature points. The orientation and the scale are also computed using the voting. Finally, the algorithm clips ROI area from the representative point. This algorithm replaces a template image by the ROI area. We verify experimentally that this algorithm correctly track rigid body motion.
