Rapid and high-accuracy guidance line identification and tracking are the keys to ensure AGV’s real-time control performance. Analyzing outdoor environment under different lighting conditions, we proposed a feature extraction method which could guarantee the efficiency and stability of line recognition for different surroundings; analyzing how angular deviation and distance deviation affect AGV posture, we designed a subdivision-control tracking strategy. The results show that the algorithm has achieved effective identification and tracking with an error less than ±5cm.