부산대학교 도서관

Autonomous buses are expected to address traffic and social issues. Safety of passengers is an extremely critical issue which needs be considered when implementing autonomous driving services. Passengers can fall when travelling on the bus especially in situations such as sudden braking etc. This is a very dangerous situation which needs to be handled immediately especially in the case of elderly people who are more prone to injuries. To manage such situations, an autonomous bus without a driver requires a remote supervisor who is responsible for monitoring the cabins of multiple buses through cameras. In this paper, we discuss a method for detecting falls inside the bus by using image processing technology, which assists the remote supervisor. During the experiment, 90% accuracy was achieved in detecting falls inside the bus, and we discussed a method to detect falls with robustness in implementation in the actual scenario.