Aiming at the guidance path recognition technology based on vision guided automatic guided vehicle (AGV) system, it is vulnerable to the influence of light intensity, pavement reflection, shadow and other factors. To solve this problem, a method of image dynamic threshold acquisition based on path prediction is proposed. By identifying the path point, using linear interpolation income function estimates the current path point image region. Within this area find the optimal threshold. According to the thresholds of the image binarization, find out the current region of the guide line, according to the guide line of its own characteristics to determine the threshold is valid. To get to adapt to the current environment of threshold, and according to this threshold to extract the guide wire. Through this method, the path information is collected, identified and processed, so that the AGV can be faster and more stable on the runway.