Key-Point Feature Detection Method for Surrounding-Field-of-View Image Applications

摘要

This paper describes a robust feature detection method for omnidirectional image which used for target and region-of-interest (ROT) detection. The omnni-directional system can provide much larger FOV (field of view) which can support 360° of the whole environment. Firstly, we briefly introduce the back-transformation model of the omni-directional system. Then, the Harris and SIFT (scale invariant feature transform) is applied to find the key-point features of the around area. Finally, we compared the above feature detection methods (Harris and SIFT) for the omni-directional image and its corresponding unwrapping panoramic-cylindrical and solve this matching problem with some experiment and discussion.