Dr. Nazrul Islam of ODU, Video Pattern Recognition
Thursday, 20 March 2008 at 6:30 PM
Aberdeen Barn, Virginia Beach
Automatic identification of a specific object or pattern in an arbitrary input scene is an important part of any authorization, monitoring and security system. Pattern recognition is always a challenging issue because the targets are often non-cooperative, the scene may contain noise and distortions due to variable environmental conditions during recording the image. Additional requirements for an efficient pattern recognition system are that the architecture should be simple so that it can easily be implemented and be user friendly, and it should perform fast enough to make instantaneous decision on the presence of a target in the input scene. Optical joint transform correlation (JTC) technique has been found to be a versatile tool for real-time pattern recognition applications, which employs optical devices, like lens, spatial light modulator, for parallel processing of the given images. The JTC scheme provides a number of advantages over other correlation techniques, like Vanderlugt filter, in that it allows real-time updating of the reference image, permits parallel Fourier transformation of the reference image and input scene, operates at video frame rates and eliminates the precise positioning requirement of a complex matched filter in the Fourier plane. Several modifications have been proposed to improve the correlation performance of the JTC technique, namely binary JTC, phase-only JTC, fringe-adjusted JTC and shifted phase-encoded fringe-adjusted JTC. This presentation will review the features, problems and prospects of optical pattern recognition techniques.
Posts
