A parallel algorithm for visual tracking of multiple free-swimming robot fishes based on color information

Abstract

Obtaining information of both action of multiple free-swimming robot fishes and environment quickly and accurately from image sequence is the foundation of making decision and control in a MRFS (multiple robot fishes cooperation system). The real-time vision subsystem of MRFS and its tracking strategy for multi-target are introduced in this paper. Combining features of the robot fish and location background, an adaptive segmentation algorithm based on hue histogram and saturation histogram is proposed, which can adapt the changing environmental conditions. At the same time, integrating with computer parallel processing technology, the overall tracking algorithms are optimized by means of MMX and SSE instructions. The designed visual subsystem has been implemented in MRFS, and the results have shown its effectiveness by successfully tracking multiple free-swimming fishes and obstacles.