Speedy detection algorithm of underwater moving targets based on image sequence

A real-time algorithm is presented to detect moving targets in underwater complex environment. The difference image is obtained by subtraction operation in frames of image sequence. An appropriate threshold is selected and the binary image of the difference image is obtained. Every part of the binary image is segmented into one single image by image segment algorithm. A definition of linear level is given based on the shape of objects formed by motion of targets in the difference image. The actual sizes of the original moving targets are inferred from linear level and area of the formed objects. The function of real-time detecting underwater moving targets is implemented using this algorithm. The algorithm has been used in underwater intelligent monitoring system in some large international games and meets the requirement of real-time. In practice, it is proved a fast and stable algorithm.