Real Time Earthquake's Survivor Detection using a Miniaturized LWIR Camera

In this paper a system suitable to perform precise and fast earthquake's survivor detection using a miniaturized Long Wave Infrared (LWIR) camera is described. Main challenge of this work is the detection environment which may be characterized by smoke, dust, rubble and extremely narrow spaces as well as the extremely low resolution of the continuous moving LWIR camera. To this direction the thermal information received by the LWIR camera is exploited. In addition, research is carried out in order to implement feature descriptors for detecting only parts of partly occluded people (arms, etc.) in order to reduce false positive ratio. The proposed system achieves real time earthquake's survivor detection using a miniaturized LWIR camera. The results have been tested and evaluated in real life conditions using two different LWIR cameras for proving the robustness and the accuracy of the developed system.