Color model selection for underwater object recognition

Abstract

In this paper, the characteristics of underwater image and the importance of color features in underwater object recognition are presented. Classical illumination invariant color models are analyzed. Our goal is to study and evaluate the various color models in underwater object recognition applications. The illumination invariant color models yuv, c1c2c3, HSy, rSc2 and uSb are applied to carry out comparison experimental tests and the traditional primary colors of underwater targets are selected as object identification. The experiment object images with different illumination intensities are captured by a color video camera equipped in the underwater robotics. The comparison experimental results demonstrate that the yuv and uSb color model achieve high ART (Average Recognition Rate) and lower ERT (Error Recognition Rate).