导　　师： 段发阶

学科专业： H0402

授予学位： 硕士

作　　者： ;

机构地区： 天津大学

摘　　要： 智能相机是近年来机器视觉发展的一个新方向,其所具有的潜力注定智能相机将是未来视觉检测方案的主流。然而在国内,智能相机的发展方兴未艾,关键技术基本处于探索阶段,同国外的大公司,如Cognex,Matrox,Basler等相比,国内为数不多的智能相机研发商,如大恒,凌云,视觉龙,嘉恒中自等都鲜有自己的智能相机产品,更多的是代理推销国外的产品,市场也基本被国外公司所垄断。本文在研究了智能相机国内外的发展状况下,综合前期的研究成果,在充分研究了目前的两款主流处理芯片DSP和FPGA做数字信号处理所具有的各自特点的基础上,提出了一种基于DSP+FPGA相融合图像处理技术的智能相机方案,并设计了系统原理图,制作了样机。针对图像预处理中应用极为广泛的邻域算法,详细地讲解了其FPGA实现的数据流程控制方法,并以具体的sobel边缘检测算法为例,给出了其检测的实验效果图。针对芯片定位的应用,在实验室条件下搭建了二维机械臂模拟系统,并对系统进行了静态标定。同时深入研究了芯片定位的图像预处理算法和高级检测算法,并通过实验对系统的误差进行了分析。 本文的主要工作内容如下: /(1/)研究了机器视觉国内外的发展状况,分析了作为机器视觉两大支柱的在线视觉检测系统和智能相机所具有的各自特点,总结了智能相机的优势,进而深入分析了智能相机实现的关键技术要求。 /(2/)在实验室前期研究成果的基础上,研究了图像处理算法的特点,分析了现有视觉方案的优缺点。在充分研究了DSP和FPGA所具有的各自特点情况下,提出了一种基于DSP和FPGA分层图像处理技术的视觉方案,并设计了系统原理图,制作了样机。 /(3/)针对图像预处理中应用极为广泛的邻域算法,编程实现了其FPGA数据流程控制,并针对具体的sobel边缘检测算法,详细介绍了每个模块的功能和状态图,并给出了实验检测的效果图,同时对这种图像数据流的控制方法进行了速度测试和评估。 /(4/)研究了FPGA的多级图像预处理方法,分析了按行处理的图像处理算法的局限性。针对芯片定位的应用,研究了其图像预处理和高级图像处理算法,并通过Matlab、OpenCV等工具对算法进行了仿真实验。研究了DSP和FPGA多级流水线结构。 /(5/)在实验室条件下搭建了芯片定位机械臂模拟系统,并对系统坐标系进行了静态标定,在实验的基础上对系统的误差进行了分析。 Nowadays, smart camera has become a new orientation of machine vision. Its potential of detecting on line made it the mainstream way of the future vision measurement. However, the key technology of smart camera has not been fully developed in China and the technology is still at a low level. Compared with the international cooperation like Cognex, Matrox and Basler, some vision cooperation in china like DaHeng, Lingyun, Vision Dragon and JiaHenZhongZi don’t have many patented products and can’t dominate the home market. Based on the research of the development of smart camera and consideration of the characteristic of DSP and FPGA, this article has proposed a principal of smart camera based on DSP and FPGA hierarchical image process technology. Further more, the schematic and prototype have been designed. The control of data flow in FPGA has been explained in detail, and the widely used neighbor arithmetic has been programmed in FPGA. And the sobel arithmetic has been explained as an example. A simulative machine arm used for chip location has been set up at the laboratory, and the coordinate system was calibrated in static. The arithmetic used for chip location has been researched deeply, and the error in the measurement has been evaluated based on the experiments. The main work is shown as follows: /(1/) Based on the research of the development of machine vision home and abroad, this article has analyzed the separate characteristics of the two main kind of machine vision–vision based on PC and smart camera, and has summarized the advantages of smart cameras, and then the key technology of the smart camera has been studied. /(2/) Based on the previous achievement of vision sensor, the feature of image process arithmetic is researched, and the characteristics of the existing vision system is analyzed. Fully considering that DSP and FPGA have their own features while processing images, a sort of hierarchical image process technology based on both DSP and FPGA has been proposed in this article. The schematic and prototype have been designed. /(3/) Considering of the widely use of neighbor image process arithmetic, we implement the control of data flow in the program based on FPGA, As an example, the diagram of the program of sobel arithmetic has been explained in detail, and the result has been shown in the article, in additional, the efficiency of this method has been evaluated. /(4/) Principal of Multi-level image process by FPGA has been explained, the shortcomings of image process based on row was analyzed. The arithmetic used for chip location has been researched, and simulation experiments of the arithmetic have been made by the tools of Matlab and OpenCV. Image process pipeline based on DSP and FPGA has been researched. /(5/) A simulative machine arm used for chip location has been set up at the laboratory, and the coordinate system was calibrated in static. The error in the measurement has been evaluated based on the experiments.

关 键 词： 机器视觉 智能相机 图像分层处理 邻域算法 边缘检测 芯片定位

分 类 号： [TE]

领　　域： [石油与天然气工程]