PCBs故障红外诊断系统中基于遗传算法的图像处理研究

【作者】 郑华；

【导师】 叶玉堂；

【作者基本信息】 电子科技大学 ， 光学工程， 2003， 硕士

【摘要】 随着红外热成像技术的发展，基于红外成像检测温度场和图像分析相结合的红外诊断技术，在电力、电子、石油化工、医学等领域，得到了广泛应用。目前，印制电路板（PCBs）故障的诊断已经成为一个具有重要意义的研究方向,传统的PCBs自动测试技术受到了前所未有的挑战。与常规方法相比较，红外诊断具有非接触性的特点，能快速诊断、隔离PCBs的热模式故障，大大节约了费用开支，减少了工作量。论文根据项目组课题“PCBs故障红外故障诊断系统研究”，针对系统中图像处理部分的图像特征提取，图像配准等关键性问题进行了深入研究。论文在红外辐射理论的基础上引出运用红外热成像技术诊断PCBs故障的基本原理；研究了PCBs故障的热模式，分析了红外热成像技术对PCBs热模式故障的诊断能力;研究了数字图像处理技术在红外诊断中的应用，并对PCBs红外图像运用Roberts、Sobel、Prewitt、Laplacian、Canny等方法进行了边缘检测;分析、建立了图像配准的数学模型，将图像配准问题转化为基于遗传算法的最优配准参数的寻优问题，并运用遗传算法理论与技术对红外图像的配准问题进行了遗传算法设计。论文中解决问题的方法，具有一定的新意。在遗传算法理论与技术基础上，深入研究了图像配准问题的编码、译码、适应度评价以及选择、交叉、变异算子的设计等。算法设计上体现了一定的新意，并对算法的优劣及其针对图像配准问题的可行性和有效性进行了分析与验证。最后论文设计了PCBs故障红外诊断系统中基于遗传算法的图像处理仿真系统，包括硬件组成、软件设计等。仿真实验运用边缘检测方法提取了图像的主要特征，然后在图像边缘特征的基础上，对图像运用遗传算法进行了配准，获得了较为理想的配准参数，并根据配准参数，对图像进行了减影运算，图像减影结果准确定位了故障部位，仿真实现了对PCBs故障的诊断，验证了遗传算法用于图像配准问题的可行性和有效性。论文对PCBs故障红外诊断系统中图像的处理进行了研究，获得了较为理想的仿真结果，并对算法设计提出了改进措施与方法，其中的研究方法具有重要<WP=7>意义和一定的参考价值，为进一步的研究工作奠定了基础更多还原

【Abstract】 With the development of Infrared Imaging Technology(IIT), the infrared diagnosing technology, based on infrared inspection of thermo-field and image analyzing, is applied more and more in field of electric power, electronics, petrochemistry, medicine, etc.. At present, the diagnosis of PCBs fault is one of diagnosis field with exploratory significance. Conventional automatic testing method come up against unprecedented challenge, compared with Automatic Test System(ATS), the advantages of infrared diagnosing technology are remote inspecting, capable of quickly diagnosing and isolating thermal fault, and laborsaving. In this thesis, according to project: the study of infrared diagnosing system of PCBs fault, we studied some key technologies of image processing deeply, such as edge inspection and image registration.Based on fundamental Infrared Radiation(IR) theory, the infrared diagnosing technology is derived of. In this thesis, the thermal pattern of PCBs and the diagnosing ability of IIT are analyzed. The digital image processing technology and its application in infrared diagnosing of PCBs fault is studied, and the edge inspection of image is given by Roberts, Sobel, Prewitt, Laplacian, Canny methods. The model of image registration is analyzed and established, an infrared image registration method for diagnosing of PCBs fault based on genetic algorithm (GA) is proposed, and the algorithm is designed. Some new ideas are embodied in this thesis.Based on GA theory and technology, for the image registration, the encoding, decoding, fitness evaluation, and the design of selection operator, crossover operator, mutation operator are studied deeply, some new ideas are embodied in the algorithm, the advantages, inferior position, feasibility and efficiency of the algorithm are analyzed particularly. At last, the simulating experimental system of image processing is built, include hardware, software, etc..During the simulating experiments, first, the edge of image is inspected; next, the standard image and the test image are registrated by the algorithm procedure<WP=9>based on GA, at last, the images subtraction is used to locate the fault elements, and some preferable results are obtained. The diagnosis of PCBs fault is realized by simulating system, and the feasibility and efficiency of the algorithm are validated.The investigation on image processing of the infrared diagnosing system of PCBs fault is implemented theoretically and simulating experimentally, and some preferable results are obtained. The theoretical limits and the difficulties of algorithm are discussed, and the measures to improve the efficiency of algorithm are suggested. The works here are essential for further research on the study of infrared diagnosing system of PCBs fault更多还原