Abstract: In view of characteristics of high noise, amounts of moving targets and target crisscross of video sequence in coal mine underground, an algorithm for moving targets tracking in coal mine underground based on layered optical flow was proposed. Firstly, the initial target tracking templates was determined by using the difference method between two adjacent video sequence frames, and then the feature points of the template was extracted by SUSAN corner detection. Finally, target detection and tracking was realized by improved layered optical flow algorithm, and the template was update in the process of tracking. The algorithm realizes stable tracking of multi-object moving object in coal mine underground, and can solve the problem of partial occlusion target. The experimental and simulation results confirm the effectiveness of the algorithm.