图像处理在定量金相分析中的应用

【作者】 肖海波；

【导师】 赵鸿金；

【作者基本信息】 江西理工大学 ， 计算机应用技术， 2008， 硕士

【摘要】 20世纪70年代以来,半固态成形技术发展迅速,取得了很大的进展,但是半固态金相组织的分析和计算都是传统的定性和半定量的方法。随着定量金相和体视学的发展,人们也在研究材料显微组织与成分及性能之间的有规律的函数关系。但是在确定这种规律的过程中,需要测定金相组织的特征参数。如果依靠手工的方法进行测定不仅计算结果不够准确,而且测定过程繁琐,工作量巨大。近年来,随着计算机图像处理技术和定量金相的发展,基于图像处理的定量金相分析引起了广泛的关注。本文较全面地研究了图像处理的理论及方法和金相分析技术,以实现半固态镁合金的定量金相分析为目标,利用MATLAB编程,设计完成了一个界面友好,使用方便的金相分析系统。本论文主要研究内容及成果表现在:1.对半固态灰度图像进行分析,利用常用的阈值化法、FCM聚类法和数学形态学算法对金相图像进行分割,分析了分割效果。提出了一种基于金相图像三维直方图的FCM分割算法及其改进的模糊核聚类KFCM算法,提高了分割效率,并得到了好的分割效果。2.利用数学形态学知识,针对数学形态学单尺度结构元素的局限性,提出了基于数学形态学的多尺度开运算,并有很好的分割半固态金相图像效果。3.对分割后的半固态金相图像,统计出了图像中各个目标物的个数,提取出了图像中目标晶粒的面积、周长、形状因子等组织特征参数。4.采用面积法、截线法对半固态金相进行晶粒度评级;用MATLAB实现系统的集成,完成特征参数的保存。更多还原

【Abstract】 The technology of semisolid metal processing moved forward fast and had maken a great progress since 70s of 20th century.But people analyzed and appraised metallographic structure of semisolid metal by using traditional quantitative and half-quantitative methods.Recently, with the development of quantitative metallography and stereology,the regular functional relation between the microscopic structure of metal material and its performance begined to be studied.To get the regulation,some characteristic parameters of metallographic structure need be measured.And this task couldn’t been done only by handwork because of its inaccurate calculating results and heavy work.With the development of quantitative metallography and technology of digital image processing, The quantitative analyses of metallography based on digital image processing attracted many researchers and become a very active field.This paper studied the current latest advancement in metallographic analysis field and digital image processing.In order to automatically measure semisolid metallographic characteristic parameters,we built a system based on MATLAB with a friend and easy-using GUI.The main contribution of this thesis was as follows:1. Analyzed gray image of semisolid, using some segmentation methods to segment the microstructure image, such as threshold method, mathematic morphology method. And analyzed the segmentation results. Taking into account the merit of three-dimension histogram, proposed a fuzzy cluster algorithm(FCM) to segment the image based on three-dimensional histogram, and a kernel-based fuzzy cluster algorithm(KFCM), which was proved to be superior to FCM.2. Aiming at the limitation of single-scale algorithm of mathematic morphology operator, proposed a multi-scale algorithm of mathematic morphology open operator to realize the semisolid image segmentation with a satisfactory result.3. Got the result of the features of the semisolid macrostructure,such as area,perimeter,and shape factor of every single particle in the segmented image.4. Based on the method of intercept and area,realized the measurement of grain fineness of semisolid metal. Integrated image analysis system, and saved the results of feature parameters.更多还原