In this paper, we proposed a method of detecting obstacle from image sequences of ultra telephoto lens camera mounted on train. This camera was necessary to monitor a great distance 600 or more meters over brake distance. First of all railway track environment recognition that is fundamental problem in applying image processing to railway is discussed. Our method detects a moving object, and tracks it from the optical flow data. We can compensate the flow filed caused by the camera vibration, so our methods works irrespective of shake of the train. The experimental result using real image sequences proves the effectiveness of our proposed method.

Copyright (c) 2003 Railway Technical Research Institute