A novel method of predictive collision risk area estimation for proactive pedestrian accident prevention system in urban surveillance infrastructure

Road traffic accidents, especially vehicle–pedestrian collisions in crosswalk, globally pose a severe threat to human lives and have become leading cause of premature deaths. In order protect such vulnerable road users from collisions, it is necessary handle possible conflict advance warn users, not post-facto. A breakthrough for proactively preventing pedestrian recognize pedestrian’s potential risks based on vision sensors as CCTVs. this study, we propose predictive collision risk area estimation system at unsignalized crosswalks. The proposed applied trajectories vehicles pedestrians video footage after preprocessing, then predicted their by using deep LSTM networks. With use trajectories, can infer areas statistically, further severity levels divided danger, warning, caution. validate the feasibility applicability system, assessed two spots with different mobility environment Osan City, Republic Korea. As result, ratio dangerous scenes higher Spot (0.115) than B (0.077) when applying best performance model each spot, found that situation varies depending environment.