Real-time accurate egomotion estimation for trains using optical flow The accurate estimation of the speed of a train is of critical importance for security and efficiency issues in railway transportation. The traction power is directly affected by the correct estimation of the actual speed of the train, specially in transitions from stopping to moving situations. Traditional sensors are costly and some of them (like wheel encoders) can provide very erroneous measurements in adverse weather conditions due to sliperage, etc. In Vicomtech we are developing a vision-based system that uses optical flow to determine the egomotion of the camera installed in the train. As shown in the videos, the method is based on the determination of the apparent motion of point-wise features of the image from one frame to the next one in the sequence. That information can lead to the computation of the motion of the camera by means of projective geometry primitives.