Indoor trapped-victim detection system

Large scale disasters usually caused people trapped inside a building. In many cases, these people lost their motion ability and consciousness. For example, after a strong earthquake struck, people might first be hit by heavy objects or forced down by dropped objects, and then lost consciousness. The situation becomes dangerous if trapped people cannot be detected and pull out in time. Therefore, finding trapped victims inside a building in time becomes a very important issue for responders and command centers. Due to the popularity of video surveillance systems, several image-based methods have been proposed to detect trapped people by using machine learning technology to identify human features. However, trapped victims were usually covered by fallen objects. Detecting human features of trapped victims could be difficult. In this paper, we proposed, Indoor Trapped-Victim Detection System (iTEDS), an image recognition system to identify and recognize trapped victims inside a building. Given a video, we track people and big objects, such as bookcases, tables, chests and so on, inside a building. If a huge object falls on top of a person and makes the person unable to move for a while, we mark the person as a trapped victim. In order to investigate the effectiveness of iTEDS, we applied iTEDS to the accident that a dresser fell on twin boys. Our results showed that iTEDS successfully identified the trapped boys underneath the dresser.