中文摘要：

为了对金相组织的多相组织含量进行定量分析,提出了基于学习模式的金相组织识别算法对封闭的金相组织区域进行定量分析。首先,通过Canny算法和灰度等高线相结合的算法获取金相组织的封闭边缘;然后,对封闭边缘所形成的金相组织分割区域进行特征提取;最后通过软件系统提供的人机交互方式学习用户对具体单元区域的定义,据此形成分类特征描述,并将该类描述应用到未分类单元区域的分类上,达到对所有的单元区域进行分类识别的目的。实验结果表明：采用该系统进行的金相组织含量测量误差可以降低到±1%,可测组织含量的范围达到5%~95%,能满足多相金相组织含量测定等定量分析的要求,且普遍适用于不同的金相组织计算。

英文摘要：

In order to quantitatively analyze the polyphase structure contents of metallographic structure,this paper presents a metallographic structure recognition algorithm based on learning pattern to analyze the enclosed metallographic structure.Firstly,the enclosed edge of metallographic structure is acquired using Canny algorithm and grayscale contour line.Then,the features of the region surrounded by the enclosed edge are extracted.Through human-computer interaction,the system learns the definition of the user selected specific region to form the classification rules about the features.Lastly,other regions are classified according to these rules.Experiment results indicate that the measurement error of the polyphase structure contents of the metallographic structure is decreased to ±1%,the measurement range of different sections of one metallographic image is 5%~95%;and the method is suitable for different shaped metallographic structures.