Abstract
We develop a new image processing method to detect obstacles that are
moving quickly. The key point of proposed method is the motion vector generation. We apply
high frequent sequential images captured by a high-speed camera to generate motion vectors.
This approach enables us to get motion detection with simple algorithm.
Algorithm has 4 steps which are "Abstract features of objects", "Normalizing the feature" ,
"Measuring holding time of normalized features " and "Calculating velocity of moving
objects" . Quick detection of moving objects is easily performed with a lower computer
performance. The miniature model based simulator is built. In addition, we validate the
performance of our proposed method by checking some traffic scenes in the simulator.
Finally, we have experiments of the detection of objects that is trying to cross the road
suddenly. In this situation, target, moving vehicle is detected within 0.2 sec. We realize the
proposed algorithm provides the foundation for constructing a practical driving assistance
system.
Keywords - Obstacle detection, High frame-rate camera, Motion vector, Optical flow