Deep learning for subway pedestrian forecast based on node camera

Compared with other scenes, pedestrian detection in a subway station has the characteristics of large traffic flow and a high degree of occlusion. To address this issue, a detection scheme based on SSD deep network is proposed in this paper. By comparing a variety of deep networks, the head and shoulders pattern is introduced, and a deep learning network training mode for subway monitoring network is achieved combined with the subway station environment. Based on deep learning, this paper presents a subway pedestrian flow detection algorithm and designs a detection system of pedestrian flow within the subway station. The proposed system uses cameras in a subway station to collect the image information at key nodes, and predict the total number of people at the subway station after image processing. The method based on deep learning is used for detecting pedestrians in the vertical direction, which can effectively solve the occlusion problem of pedestrians. The experimental results show that the proposed method can improve the detection performance and real-time performance of pedestrian detection in subway surveillance, and can also accurately and rapidly locate a single pedestrian with a higher detection accuracy.