High-Speed Intelligent Surveillance System for Security Applications

Abstract

In recent times, strengthening of security through surveillance techniques has gained significance. In highly secure places, like military reserve zones and airports, surveillance is still being done manually. Though cameras are deployed in surveillance, its function is limited to recording for the follow-up investigations. There is no built-in face recognition mechanism in the cameras. There are possible errors due to human negligence. While face detection algorithms are deployed, the system has to be trained for the whole dataset whenever new data are added. This consumes more time and increases the hardware complexity, which renders the system unsuitable for real-time applications.Therefore, this paper proposes an effective facial recognition based on the correlation of input image data from the camera and the database in Fourier domain. The face is detected based on the peak of correlation. Even covering of different facial parts, namely, eyes, nose, mouth and tonsured and shaved faces can be predicted by correlation. Therefore, FFT based face recognition system is efficient. The video input is fed as distinct frames for correlation, in Fourier domain, with the database. A peak in correlation indicates that the input image matches with the image in the database. Hence, the surveillance process can be automated using face detection capabilities to a DSP platform and it is less time consuming than the conventional machine learning algorithms.