In order to obtain better robustness and real-time processing of path recognition, the binary segmentation image, noise searching and filtering procedures were employed. The whole path was classified into far part and near part, and the difference between the two parts was used as a threshold for selecting target road. So the navigating parameters were adapted to the changing path. Based on a model of four-wheel intelligent vehicle, a path tracking simulation was performed. Then, the navigation system based on path tracking was designed by using two digital signal processors. The experiments showed the accuracy and robustness of the system.