Fast stereo matching for autonomous vehicle navigation

The established approach of stereo vision is to first compute the disparity image by establishing correspondences across the input images and then to compute the 3D of every pixel in world coordinate system. However, it meets many difficulties when it is applied in close navigation for autonomous vehicle. This paper proposes a universal binocular stereo vision method based on linear camera model. It could handle image data taken from arbitrarily positioned cameras. Furthermore, the image rectification process that usually is difficult and time-cost is unnecessary in our method. Based on it, we present a fast and robust stereo algorithm that is applicable to close navigation for autonomous vehicle. In our algorithm, height estimation is used to decrease the search range of matching and height image is obtained directly. Experiment results with real image data are presented to illustrate the performance of our algorithm.