In this paper, an automobile collision avoidance system based on video image processing is proposed for vehicle active safety. Firstly, considering the requirement of real-time system, FPGA is responsible for the core task of video image processing. In the process, this paper introduces an image segmentation algorithm to extract the Region of Interest, a fast median filtering algorithm and an improved adaptive threshold Ostu segmentation algorithm. Secondly, in the design of alarm system, we use the STM32 chip which is based on Cortex-M3 kernel as the embedded platform. Finally, in order to analyze the data obtained by image processing and make smart decisions on the embedded platform, we establish a double parameter alarm mechanism of automobile security model. It can help the drivers to take appropriate hedging measures. The system in this paper has the characteristics of low cost, real-time and high reliability. Through the debugging and experiments, the system has reached the expected effect and met the requirements of automobile collision avoidance system.