A New Feature Extraction Method for Underwater Targets

Abstract

in the complex underwater environment, underwater images are taken by special underwater CCD camera and its S/N is low and the edge is fuzzy. For the four types of characteristic underwater targets, the novel moments called relative boundary moments are proposed, and the affine invariants of discrete moments are constructed. With scale, translating and rotating invariance, the moments can be used as the descriptors of the samples. Experimental results show that compared with the traditional regional moments, the new moment invariants not only can reduce the calculation in data processing to a large extent, but also improve the robustness and timeliness for engineering applications. When applying to the practical engineering, that is particularly approval for AUV to complete a certain mission.