One Fish, Two Fish, Butterfish, Trumpeter: Recognizing Fish in Underwater Video

This paper presents a deformable template object recognition method for classifying fish species in underwater video. This method can be a component of a system that automatically identifies fish by species, improving upon previous works which only detect and track fish and those that rely on significant inter-species shape variations or special equipment. Our method works with video shot by a standard uncalibrated camera in a natural setting rather than the calibrated stereo cameras and man-made imaging environments described in other publications. We use deformable template matching which employs an efficient combination of shape contexts and large-scale spatial structure preservation. Experimental results demonstrate the improvement of deformable template matching over raw SVM texturebased classification.