Multisensor of thermal and visual images to detect concealed weapon using harmony search image fusion approach

Perils of terror threats are a problem that can be managed by precise and competent surveillance systems. It is an imperative demand to swiftly conduct real time verification of person suspected to borne threat. Real time verification exercise involves identification of concealed weapon threat, interrogation of suspect and seizure of weapons or explosives in a covert controlled fashion. There is an urgent need for an intelligent and safe surveillance system with inbuilt sensors that can detect concealed weapons and pinpoint their location on the body without any physical contact from a standoff distance. Fused imagery of visual and infrared images is processed to identify weapons specifically. When the weapon template is matched with the suspected object using Zernike moments, the identity of the object becomes obvious, and any effort to conceal it becomes futile. In this paper, a novel approach to detect concealed weapons based on discrete wavelet transform in conjunction with dimension reduced meta-heuristic algorithm, the harmony search, and shape matching based K means SVM classification is presented. Experimental results are provided to demonstrate that the proposed hybrid approach provides superior performance.