Abstract: Distortion and mass data are difficult problems in image sequence mosaic on vehicle-borne camera platform. A fast image sequence mosaic method was proposed for solving the problems. Firstly, the distortion was removed after image preprocessing. Secondly, the speeded up robust features (SURF) were extracted from preprocessing images and the features were matched. The rate of feature matching was defined as a similarity measure of frames to extract key frames. Then, by using the improved image registration method without accumulated error, a global model was established. Finally, the image sequence mosaic was obtained through image fusion of the maximum. The experimental results show that the method is robust and fast.