An embedded system for disaster detection and rescue based on an improved CAMSHIFT framework

Fire accidents and some other nature disasters have been great threats to the human society. However, there's no uniform framework for detecting and monitoring such disasters. Therefore, we proposed a prototype embedded system for early fire detection and monitoring based on computer vision techniques, aiming at high reliability and low cost. By employing a bidirectional histogram training method and an extended CAMSHIFT tracking framework, the proposed fire detection system can detect the suspicious fire region accurately and robustly with very low hardware costs. This framework can be conveniently extended to some disaster detection and rescue projects like the search for survivors from the shipwrecks. Experimental results have proved our proposed system a promising one for massive deployment and daily usage.