Pedestrian Detection in Metro Station Based on Improved SSD
Xiaowei Dong, Yue Han, Wei Li, Bo Li · 2019
In crowded scenes such as subway stations, due to the high pedestrian density and high occlusion, most existing pedestrian detection methods are difficult to obtain a good result, and have poor robustness to small targets. In response to this problem, our work proposes a modified human detection network which based the SSD(Single Shot MultiBox Detector) model. First, combined with the pedestrian data analysis of the scene, the head and shoulder model is introduced. Then, the SF-SSD network is proposed. Based on the SSD network, the low-level feature map is merged and its location information is used to accurately detect small targets. Finally, experiments are carried out on the subway pedestrian dataset. The experimental results show that the proposed method of SF-SSD network effectively reduces the false positive rate and missed detection rate of SSD network on subway pedestrian dataset, and meets the real-time performance requirement of pedestrian detection. At the same time, the robustness is improved, and pedestrian detection in dense crowd scenarios is effectively realized.