Aiming at short comings of low precision and poor real-time of traditional odometer

feature point extraction of stereoscopic image sequence is performed using template feature point extraction algorithm combines Blob with Corner. Combine improved sum of absolute differences（ SAD） with ring matching model for feature points to achieve exact matching of feature points. Distance and direction information of vehicle movement are obtained accurately

by geometric model for pose parameter estimation and RANSAC iteration of three-point algorithm. The experimental results show that the stereo vision odometer based on feature point extraction of template has high precision and good real-time performance

which can provide a new way for vehicle auto navigation.