The Research Of Image Processing In Moving Object Detection And Recognition By Using Joint Transform Correlator

Conventional matched filter-based correlator needs to make matched filter in advance, it has many disadvantages: large calculation, slow speed, not suitable for real-time processing, and recognition precision is low. So optical joint transform correlation is proposed. Optical joint transform correlation can recognize automatically, overcome the drawback of filter-based correlator, and it is suitable for real-time processing, can achieve better recognition result. However, the correlation signal obtained by the classical matched filter is associated with large correlation sidelobes, largecorrelation bandwidth, and low optical efficiency.In the paper, we describe the whole configuration of the joint transform correlator; introduce the correlative theories of joint Fourier transform in detail; In order to achieve better correlation result, the nonlinear processing of joint transform power spectrum (JTPS) is put forward, the effect of JTPS Laplace sharpening and binary processing to the correlation image is given; the method of the best binary threshold is determined; we analyse the correlation image, present the way that removes liquid crystal diffraction spot, direct current spot and image noise, and binarize the correlation image finally.