Joint Transform Correlator In Research On Image Recognition

With the development of modern technology, more and more researchers in the information processing field realized the importance of image pattern recognition. Optical systems, which have the natural 2-D parallelism, provide a basement for the image recognition field. The correlators have been paid more attention to in the recognition field. The NSF, DARPA and USAF had held many special forums discussing the optical corrrelator抯 utilization in pattern recognition. A non-real time recognizer had been produced in our country, but it is based on the correlation-peaks?threshold value, so the error maybe appear. Furthermore, the JTC seriously depends on the characteristics of signals and noise, and the noise has heavily influences on the recognition results. To improve the JTC抯 recognition ability, we researched the improvements of the accuracy and reliability of the image recognition.A objects recognition method based on the normalization fringed adjusted joint transform correlation was proposed in this paper. Firstly, theory of joint transform correlator was expatiated based on its concept, then a classical joint transform correlator was analysed and its shortcomings were showed, some parameters which are used to measure the correlator抯 performance were also discussed and proposed a normalized-based fringe adjusted joint transform correlator. Theoretical analysis proved this method抯 feasibility. The arithmetic for the computerbased simulation of this joint transform correlator was designed in this paper. The simulation result showed that the normalized-based fringe adjusted joint transform correlator works very well, that it can detect the place of object correctly, also that it can recognize the object from the input image which contains the interfered similar objects. To verify the advantage of this method, we compared the proposed correlator with normlized joint transform correlator and the fringed adjusted joint transform correlator, and caculated two parameters of theirs: sharpness of peaks and discriminability. The result showed that the normalized-based fringe adjusted joint transform correlator was a kind of reliable correlator.