Detecting the person of interest using agent based video processing

This paper presents a novel solution for finding a person of interest using a mobile agent platform developed with Raspberry PI III module in indoor dynamic environments. This method is capable of extracting human figures from complex backgrounds while the platform is still moving and compare the identified human figures against the query image using features like clothing and other accessories. The proposed method processes the video near real time with limited resources on the Raspberry PI III module and communicate the findings to the central control system using a specially designed communication protocol. System is capable of identifying the person of interest with the accuracy level of more than 90%. The proposed method can either work on its own or be and be integrated with an existing CCTV systems improving its capabilities further.