UAVを用いて撮影した光学画像における人検出の精度向上

抄録

When natural disasters strike, such as tsunamis and floods, survivors are sometimes isolated in disaster-affected areas. For searching those people more quickly, UAV (Unmanned Aerial Vehicle), which can be easily operated by remote control, is believed to have the capability. In order to make the search activity using UAV more reliable and efficient, methods for detecting survivors automatically are necessary. In this study, by applying several image analysis methods to the images taken with UAV, the authors tried detecting humans from those images with high accuracy. As a result, the authors succeeded to detect them with zero to eight misdetections and little nondetections by utilizing their whole bodies, upper bodies, shadows, and line segments in the images.