基于主动轮廓模型的PCB红外图像分割-模式识别与智能系统专业论文.docxVIP

万方数据 万方数据 PCB infrared image segmentation based on active contour A Dissertation Submitted to Civil Aviation University of China For the Academic Degree of Master of Science BY ZENG Pei-pei Supervised by Associate Prof. WANG Li College of Aeronautical Automation Civil Aviation University of China May. 7th 2014 中国民航大学硕士学位论文 中国民航大学硕士学位论文 摘 要 电路板红外图像存在着噪声大，芯片发热区域交叉等特点，现有的图像分割 算法大多针对可见光图像或医学图像，难以有效的对电路板红外图像进行分割， 因此必须对传统的图像分割算法进行优化改进，使其适应电路板红外图像分割提 取任务。本文首先介绍了传统的 Snake 模型，通过对仿真结果的分析，得出传统 模型存在着轮廓深度凹陷收敛困难和初始轮廓敏感等问题，然后研究了几种比较 热门的改进方法，气球力模型，GVF 模型，CV 模型，结合电路板红外图像的特 点对这几种模型进一步改进，仿真结果表明，改进的算法对传统的 Snake 模型的 缺陷有了明显的改善，但改进的模型还存在着初始轮廓依赖人工选取和多目标提 取问题。 结合区域增长法，GVF-Snake 模型，本文提出一种多目标自适应芯片区域提 取算法，能够在无人工参与的情形下对多个芯片区域进行自动提取，算法充分利 用了电路板红外图像的发热特性，自适应的选取种子点并生长为初始轮廓曲线， 在 GVF-Snake 模型下使初始轮廓收敛到芯片核心发热区域，完成对芯片区域的 提取，仿真结果表明，初始轮廓的选取能够加快轮廓曲线的收敛速度，本文模型 能够有效提取多个芯片的发热区域，提取精度高且速度快，有一定的工程价值。 另外，本文的算法还是一种可加速的并行算法，在今后的研究中，还可以设计并 行算法对运算速度进一步优化，使得算法更具实用性。 关键词：印制电路板、红外图像分割、主动轮廓模型、GVF、区域增长法、多目 标 I ABSTRCT Printed circuit board infrared image is characterized by much noise and chip heating area crossing, the existing image segmentation algorithms is largely for visible image or medical image, which are difficult for infrared image segmentation effectively, in order to be suit to the segmentation and extraction of circuit board infrared image, the traditional image segmentation must be optimized and improved. Firstly, this paper introduces traditional Snake Model and finds out the contour problems in traditional modal through analysis of the simulation result. Next, several popular improved methods ,such as Balloon Fore Model, GVF Model and CV Model, are studied and improved for the characteristics of circuit board infrared image. The simulation results show that the defects of traditional Snake Model has been improved obviously, but there are initial contour relying on artificial selection and multiple-objective extraction problems in the improved Model. Combining Region Grow Algorithm and GVF-Snake Model, t