Human Detection and Posture Estimation with IR Thermal Images by a Drone

Abstract

Recently, multirotor drones have been used in various fields, and drones can effectively search missing people in harsh environments. Infrared (IR) thermal imaging can be used day and night to find people that cannot be detected in visible light images. In this paper, we address the detection and estimation of a person's posture in dangerous environments using a multirotor drone equipped with an IR thermal imaging camera. We detect missing people through k-means clustering and morphological operations and remove false alarms based on the size and squareness of rectangular object windows. Then, the posture of the detected object is estimated through template matching. In the experiments, standing or sitting people were detected using thermal images captured by a drone during day and at night in the mountains, and the posture was successfully estimated.