|
Along with the development of electron, computer and artificial intelligence technology, machine visual technology has became a very important orientation for studying on the detect method of steel surface’s defects. Until the end of twentieth century, a few western countries, such as America, Germany and so on, have successfully studied the defects image preprocessing system of steel surface based on machine vision, which makes great contributions to improving quality of steel. There are a few colleges which do this research in our country, and there are no practical products, which limits the quality of steel of our country.For solving the important issue of image preprocessing in the system of steel surface detect by machine visual, use the NiosⅡembedded system based on SOPC, the default image preprocessing of steel surface is established, the characteristics of embedded system, such as high performance, abundant interface, high flexibility, abundant resource, are adequately used.Firstly, this thesis introduces some simple and fast image filter algorithm and image segment algorithm which used in defects preprocessing algorithm. Comparing these algorithm by experiment, median filter is chosen to be filter algorithms, and 1D maximum between-cluster variance algorithm is chosen to be segment algorithms.Secondly, this thesis analyses function, and build the steel surface defects preprocessing hardware, then programmed the PCI interface IP core in FPGA with VHDL language, use the SDRAM controller IP core, complete the NiosⅡsystem platform for software design and experiment platform.Thirdly, based on the hardware platform and image preprocessing algorithm, the relative software programs are written in NiosⅡIDE, complete the image filter algorithm and image segment algorithm. In order to do further image processing, configuration the PCI interface chip and program relative driver by WinDriver realized communication between hardware and computer.Finally, introduced the hardware platform of defects preprocessing system. The experiment of the image processing of steel surface is achieved, and the result of the experiment verifies the validity of the system.
|