Study to real-time target recognizing system based on optical correlator

摘要

The recognition algorithm based on geometry texture character includes pattern matching, estimation function, and determination tree. The advantage of above algorithms is easy to be programmed and realized by computer. However, the recognizing operation consumes the mass of system source, which lead to the difficult to assure the good recognizing accuracy under the high recognizing speed. Optical correlation processing system transmits information by photons and operates by photo-electronic device, and realizes the target recognition by optical refraction, interference, and diffraction. Compared with traditional computer image processing techniques, optical correlation system offer primarily two types of benefits. On the one hand, optical correlation systems have an inherent capability for parallel processing, which can transmit rapidly mass information. On the other hand, it changes the complex images processing problem into spot recognizing in output plane. The correlation peak of high gray value appeared in the correlation plane can be used to the proof to tell target from many kinds noises. The analysis in paper to how to make the complex matching filter, how to determinate the truly correlation peak from others signal in correlation plane rely on neural nets technology could be helpful to the study of optical correlation system used to the field of image recognition.