A new method of detecting passenger flow based on human head and shoulder three-dimensional information is proposed, detecting passenger flow according to spatial relationship features between head and shoulders as well as head shape, size and other characteristics, which can effectively solve system overall performance reduce problems owing to ordinary color-camera image processing being affected by light, posture, shelter and so on. With Xtion sensor as a platform, identifing and tracking human head, counting the passenger flow according to comprehensive analysis on depth image of body surface three-dimensional features. The experimental results show that this method not only improves the precision and efficiency of detection,reducing the amount of computation of data, real-time video stream can be processed, but also solves the need to establish training detection in color images. The system accuracy can be up to 96%, and has practical value.