Robot fish detection based on a combination method of three-frame-difference and background subtraction

By combining three-frame-difference method with background subtraction method, an approach is proposed for rapid detection and identification of moving targets. First the background frame is achieved after decoding frame operation on video which is collected by the robotic fish platform and made into gray-level. Then the threshold determined by the OTSU algorithm is employed to transform them into binary images. After that, a new binary image (F1) that the moving targets are extracted is obtained by three-frame-difference method mixed with AND logic operator and the mathematical morphology operation. If moving targets are detected in frame (F1), background subtraction method is applied to get another binary images (F2); otherwise specific messages will be shown in the frame. Finally, accurate image is obtained by using the 4-neighbor topology method mixed with the AND operation between F1 and F2. Program simulation experiments show that in the robotic fish system the proposed algorithm can detect fish and judge moving situation accurately, completely and quickly.