Research Of Machine Vision Based On SOC

Embedding eye and brain in machine is a dream for human beings. Machine vision, as the extension of human vision and brain, is a measure of the degree of automation of modern industry. With the rapid development of semiconductor technology, emerging technology will gradually make the dreams come true. Products with visual function have been applied in various domains, and it will promote and facilitate the intelligent development of automation. In this thesis, the implement of embedded machine vision will be researched.The SOC technology, including IP reuse and S/H co-design ,is firstly introduced, and especially, Soft core regarding SOPC technology will be concretely analyzed in this paper, such as NiosII and MicroBlaze.And then, I mainly concentrate on the introduction of CCD and CMOS image sensor, the comparison of their differences, the performance analysis of MT9V032 image sensor with global shutter. MT9V032 with a view to dynamic quality, which is more powerful than common image sensor, will be firstly used in this design.In the following paper, the mechanism of hardware will be described, including the function of all schematic module, the selection standards of controller, SRAM and power chips and their typical configurations, especially the prevention of EMI considering signal integrity and power integrity.Finally, the scheme of software will be introduced. I proved the feasibility of approach trough Matlab. The image processing IP core is designed to be compatible with the protocol of Avalon bus. It achieves the function of target identification when using ModelSim to simulate. The NiosII system with I~2C, PIO, UART and image processing IP are customized to connect. According to parallel port transfer agreement, the FPGA and PC achieve the communication. The software is compiled under the Visual Studio 2005, and the image results with high quality are given.