Pattern Recognition Using Volume Holographic Image-base

In information times, information storage technologies that can achieve both high capacity and high speed are required. Owing to the advantage in high storage capacity, fast data transfer rate, and short access time, volume holographic storage technology has been considered as the most promising, next-generation storage technology. Although the performance of volume holographic memories (VHM) has been improved very rapidly in recent years, there are many obstacles when it is used for image pattern recognition. This thesis is aiming at pattern recognition with image bases stored in VHM.This paper firstly introduced relevant fundamentals of pattern recognition, discussed pattern similarity and gave many measuring methods for pattern classification. Then the relevant volume holographic principles were briefly illustrated, followed by results of general coupled-wave theory which are useful for volume holographic storage. These paper also expressed relevant theories on pattern recognition with image-base using VHM.This paper defines the template of image recognition and characteristic recognition vector. When the diffractive efficiency of the holograms in a VHM is not uniform enough and the SNR of the retrieved images is very low, the precision of direct pattern recognition will be very low. In this case, proper error-correcting methods must be used. We have done experiments for attaining adaptive recognition algorithm and proposed the scheme of correlation coefficient (CC) as the optimum recognition algorithm. Pattern recognition for volume holographic image-base built by both angular multiplexing technology and block-oriented holographic storage technology were implemented by using CC method.In this paper, an associative memory for recognizing volume holographic imagebase was implemented. This system adopted opto-electronic hybrid feedback method. Distorted images and partial images were recognized by CC method using this system. The reference beam corresponding to the recognition result was repositioned, and replayed the original image successfully. For the associative readout of partial image, we determined the great allowed difference between the partial image and original image by measuring the correlation-peak series using CC.Through experiment investigation, this paper proposed future work for improving pattern recognition system based on VHM.