Dolphin movement direction recognition using contour-skeleton information

Abstract

Detecting and tracking social marine mammals, including dolphins, can help to explain their social dynamics, predict their behavior, and measure the impact of human interference. The underwater environment is very special and different from the land environment: acoustic recorders are the main equipment for researchers to track dolphins at a long distance. Close-range detection of visual data is often seriously affected by the underwater environment, because the underwater environment is highly dynamic and the light source has attenuation in the deep water. Nonetheless, compared with acoustic information, visual data can provide more detailed information at low cost. The videos and images of dolphins provide researchers with great convenience in studying the body structure and social behavior of dolphins. At the same time, dolphin movement direction recognition helps researchers to learn more about dolphins through a series of accurate movement data. In this paper, we proposed an approach to detect the movement direction of dolphins effectively. First, the contours and skeletons are detected, which could reduce the impact of wrong detections significantly. And then another CNN-based model obtains the movement directions with feature images extracted from the previous steps. The experiment result shows the correctness and efficiency of the proposed method.