Residual Person Monitering Method based on Deep Learning to prevent negligent Accident of Store and System thereof

摘要

The present invention provides a deep learning-based real-time residual person detecting method to prevent a safety accident in an unmanned store, which comprises the steps of: receiving video information in a store from CCTV by a video analysis management server; detecting a person based on deep learning from the received video by the video analysis management server; calculating a heat map by the video analysis management server; detecting a residual person candidate by the video analysis management server; tracking the candidate and calculating the remaining time by the video analysis management server; and determining the residual person based on the reference time by the video analysis management server.