Motion states extraction with optical flow for rat-robot automatic navigation

The real-time acquisition of precise motion states is significant and difficult for bio-robot automatic navigation. In this paper, we propose a real-time video-tracking algorithm to extract motion states of rat-robots in complex environment using optical flow. The rat-robot's motion states, including location, speed and motion trend, are acquired accurately in real time. Compared with the traditional methods based on single frame image, our algorithm using consecutive frames provides more exact and rich motion information for the automatic navigation of bio-robots. The video of the manual navigation experiments on rat-robots in eight-arm maze is applied to test this algorithm. The average computation time is 25.76 ms which is less than the speed of image acquisition. The results show that our method could extract the motion states with good performance of accuracy and time consumption.