The image sequence captured by the vehicular camera vibrates due to the vehicle vibration or harsh road conditions. It greatly affects the subsequent image processing. As a video processing technology, electronic image stabilization (EIS) can solve this problem effectively. In this paper, a gasoline model car platform is established, which has rich vibration characteristics. Besides, high unevenness roads are selected to make the vibration problem more prominent. Considering vehicle demand, feature point detection and matching algorithm is applied to match images in the process of establishing EIS. In addition, the algorithm of eliminating mismatched feature points and the adaptive Kalman filter (AKF) is improved to make them more suitable for vehicle application. The experimental results show that the EIS developed in this paper meets the real-time requirements and can work normally in harsh road conditions. Since the platform and working scenarios are more extreme than the normal conditions, the EIS developed in this paper has high portability in the environment perception system of smart car.