Using OpenMV with STM32F427 as the MCU, a machine vision-based track barrier-avoidance intelligent car system is designed. The car system can quickly look for the specified field randomly lit signal beacon from the designated location of the marking. The design uses the camera module system to collect the image, get the position information of the obstacle, and measure the distance by using the captured obstacle image further, so as to get the position signal of the obstacle. Using FAST feature point detection algorithm, NCC template matching algorithm, PID control algorithm, monovisual visual ranging algorithm, etc., when there are obstacles in front of the car they start curve exercise to avoid obstacles, when no obstacles they are straight driving, to achieve the purpose of automatic driving. The system can verify and realize the function of automatic "search" barrier by many experiments, and has the characteristics of high sensitivity and strong stability.