In this paper, the vehicle detection with distance estimation algorithm is proposed by using two CCD cameras mounted on the both side mirrors of our experimental car, TAIWAN iTS-1. The Sobel edge pixels and gray intensity are applied for the lane marking and vehicle detection in our work. A new detection which utilizes a lane prediction technique to reduce the size of the ROI is developed for real-time embedded system applications. The side vehicles can be easily detected by comparing the gray intensity with the road surface. Then, the image coordinate model is utilized for the distance estimation with the detected approaching vehicles. The verified effective distance is 30m behind the host vehicle. This approach has been successfully implemented on the DSP board for the real-time detection and this system will be extended to the lane change warning system to give drivers alerts to avoid collision.