Optical Neuro Computing System with Intelligent Sensor

1.在国际光学系统设计与制造研讨会上和欧洲的一些研究机构，对光学计算系统的发展研究进行了调查。关于多路匹配空间滤波方法的发展研究仅限于本课题组的研究以及美国和加拿大的一些研究。人们认识到，在欧洲和美洲，很少有研究达到我们的研究水平。根据激光技术在流体力学中的应用国际研讨会和欧洲一些研究机构的讨论，澄清了使用多路复用匹配空间滤波器来区分形状和尺寸的光学计算方法可以有效地应用于PIV（粒子图像测速法），用于速度可视化和瞬时同步绘制速度等值线图的方法。通过这些研究，阐明了光学反算方法识别形状和尺寸在PIV应用中存在的问题和解决方法.阐明了傅里叶-梅林滤波器可有效用于研究利用光学反演算法在旋转不变和尺寸不变的情况下识别形状和尺寸的有效方法，并且二元透镜等新光学部件的设计对于形状和尺寸识别的发展研究是必要的4。阐明了利用光学辅助计算系统结合神经网络的判别功能，对2行40个数字字母的同时瞬时识别的确定性可达95%以上。

1. Investigation of the developmental research on Optical Computing System has been done at the international symposium on Optical System Design and Production and at some research institutes in Europe. Developmental research concerning Multiplexed Matched Spatial Filtering method were limited to the research of our group and a few research in United States of America and Canada. It was recognized that few research reaching the level of our research were there in Europe and America.2. It was clarified in view of discussion at the international symposium on Application of Laser Techniques to Fluid Mechanics and at some research institutes in Europe, that the Optical Analogue Computing method using a Multiplexed Matched Spatial Filter for discriminating shapes and sizes could be effectively in applicable to PIV (Particle Image Velocimetry), for the methods visualizing velocities and making equal contour map of velocities instantaneously and simultaneously. The issue and settlement in the application of Optical Analogue Computing method recognizing shapes and sizes to PIV,were clarified through these investigations.3. It was clarified that the Fourier-Mellin filter could be used effectively for the research concerning effective methods recognizing shapes and sizes in rotation-invariant and size-invariant with Optical Analogue Computing and that the design for the new optical parts as binary lens could be necessary for developmental research for recognition of shapes and sizes4. It was clarified that the certainty for the recognition of 40 number letters in 2 rows could be done simultaneously and instantaneously for over 95% using Optical Analogue Computing System together with discriminating function with neural networks.