A Study of Parallel Micoroptical System for Image

在大规模实时图像处理领域，并行光学处理是一种处理图像信息的有力手段。提出了一种利用平面微透镜阵列的高密度图像复用能力的图像识别系统，该系统由两个主要单元组成。第一单元，即预处理器，通过使用2D沃尔什函数的核的变换从输入图像获得空间频域中的频谱。第二部分是后处理部分，利用模式匹配的方法对输入图像的空间频谱进行识别，并根据该光学处理系统进行了光学器件的研制和系统的开发。因此，获得了以下结果：1.大口径平面微透镜的制作用与透镜基底相同的玻璃制作了坩埚。然后，利用这种熔化的p ...更多信息 并结合电迁移技术，得到了直径约为1.5 mm的透镜阵列。参考滤光片的研制利用照相胶片研制了具有透明度分布的高密度参考滤光片。3.设计并制作了一对可实现光功率相减的光电探测器; 4.预处理器中谱系数的选择研究表明，选择函数的输入图像谱之间的离散度大于所有基函数的输入图像谱之间的离散度。5.对输入图像变形的耐久性;结果表明，该系统对输入图像变形（如旋转和横向移位）的耐久性强于傅立叶域中的匹配滤波。6.仿真结果与光学实验结果的比较在基本光学系统中，我们制作了该识别系统。当我们对预处理器和后处理器进行实验时，它们与模拟结果吻合得很好。然后，将两个处理器按列连接，由计算机将光图案与参考图案进行比较。这一结果也与模拟结果相吻合。少

In the large scale and real-time image processing area, a parallel optical processing is a powerful method dealing with image information. We proposed an image recognition system using planar microlenses array which has an advantage of a high density image multiplexing ability.This recognition system is composed of two major units. The first unit, namely pre-processor, obtains the spectrum in the spatial frequency domain from an input image by transformation using the kernel of the 2D Walsh function. The second unit, namely post-processor, discriminates the spatial frequency spectrum of an input image by methods of pattern matching.According to this optical processing system, we have researched to cultivate the optical devices and to develop the system. Thus, the following results were obtained : 1. Fabrication of the planar microlens which has the relatively large diameter ; We have fabricated the melting pot by a glass which is a same kind of the lens base. Then, using this melting p … More ot and the technique of the electromigration, we got the lens array which the diameters were about 1.5 mm. 2. Development of the reference filter ; We developed the high density reference filter which has a distribution of the transparency by using the photo film. 3. Design and fabrication of a pair photodetector that put out the light power subtraction ; 4. Selection of coefficients of the spectrum in the preporcessor ; It was shown that the dispersion between the spectrums of input images is larger in the case of using the selective functions than all base functions. 5. Durability against input image deformation ; It was shown that the durability of this system against input image deformation like rotations and lateral shifts is stronger than the matched filtering in the Fourier domain. 6. Comparison between the simulation and the optical experiment ; We have fabricated this recognition system in the fundamental optical system. When we experimented with the preprocessor and the postprocessor, they were good agree with the simulation. And then, these two processors were connected in column and the light patterns were compared with the reference pattern by the computer. That result was agree with the simulation too. Less