Pretreatment And Related Recognition Simulation Of Optical Image

With the development of computer science in image processing, image recognition technology becomes more and more important in life and production. It has been widely used in hand pattern recognition, industrial parts recognition, remote sensing and military target recognition, etc. For the advantages of programmable control on the computer, optical image correlation recognition has become one of the most promising methods in the real-time image recognition system. And the pretreatment and distortion of the target images are two main aspects in the efficiency of image recognition, which are discussed in this paper.In the thesis, the overall structure of the optical image recognition system is described, and the relevant principles of two commonly methods in image recognition processing are introduced. The whole process of the optical image recognition is simulated by using MATLAB of programming software. Different filter pretreatments to the input images are used, and the output correlation peaks of before and after treatment are analyzed quantitatively, in order to select the best image pre-processing filter. Meanwhile, the binary processing and power spectral subtraction processing to the power spectrum of the input images are put forward, which improve the sharpness of the first-order correlation peaks and reduce the impact of the zero-order correlation peaks on the recognition. It includes that the two image recognition correlators are compared in common use. After the synthetic discriminant fuction processing, the analysis of the output peaks is introduced. It is inferred theoretically that the target image distortion factors affect the output correlation peaks. What's more, the empirical formulas fitting the curves are come up with, which could be used for determining the distortion of other same target image.