A moving objects detection in underwater video using subtraction of the background model

This paper proposes a method for detecting moving objects on an underwater video. Video obtained using an underwater camera to capture the environmental conditions of the area. This research is the initial stage of the underwater surveillance system. Underwater surveillance system enables objects passing can be recognized shapes, types, and its behavior. The detection method used in this research is a subtraction between the current frames with the background modeling results. Underwater video retrieval has a high level of difficulty because the background is always changing either due to a change the intensity and the movement of water currents. Therefore, it needs to be made an appropriate background model to address this problem. Modeling of the background on this research using adaptive modeling method, where the intensity of the background pixels is updated based on inference of the background intensity before. If the intensity of the pixels changed drastically beyond the allowed threshold value, the pixel is considered as the pixels of the object and the pixel values of the background model are updated based on this pixel value. The effectiveness of the proposed method is expressed with the value of recall and precision. The average recall value of the two videos is 83% and the value of its precision is 67.5%.