基于改进MF-DFA特征的环形零件缺陷识别  

Component Defect Recognition Based on Improved MF-DFA

作　　者：王幸 刘文超[1,2] 陈文重 WANG Xing;LIU Wenchao;CHEN Wenzhong(School of Mechanical Engineering,Hubei Univ.of Tech.,Wuhan 430068,China;Hubei Key Laboratory of Modern Manufacture Quality Engineering,Wuhan 430068,China)

机构地区：[1]湖北工业大学机械工程学院,湖北武汉430068 [2]现代制造质量工程湖北省重点实验室,湖北武汉430068

出　　处：《湖北工业大学学报》2022年第2期11-15,共5页Journal of Hubei University of Technology

基　　金：国家自然科学基金项目(51275158)。

摘　　要：工业生产中,机械零件图像特征提取与缺陷识别通常依据零件图像的几何特征,而传统几何特征缺少零件图像的深层次特征信息作为缺陷识别特征,不适用于个性化的机械零件图像缺陷识别。多重分形是分形理论的一个重要分支,其多重分形去趋势波动分析(MF-DFA)可从整体与局部两个方面提取机械零件的深层次特征信息。联合滑动窗口、经验模态分解(EMD)与三角覆盖模型,提出一种改进MF-DFA算法,用于提取具有分形特性机械零件图像的多重分形谱特征,结合LS-SVM分类器的训练,实现缺陷零件的准确识别。实验结果表明,该算法能有效提取零件图像的多重分形特征,实现对零件图像缺陷识的准确识别。In industrial production,the feature extraction and defect recognition of mechanical component images are usually based on the geometric features of the components in the image,and there is no deep exploration of the defect recognition features of the components.It is not suitable for the image of mechanical components with nonlinear signals.As an important branch of fractal,multifractal,especially Multifractality Detrended Fluctuation Analysis(MF DFA),can deeply explore various characteristics of the target object from both the overall and local directions.This paper uses sliding window technology,Empirical Mode Decomposition(EMD)and triangular cover model to propose an improved MF DFA algorithm,and uses it for feature extraction and defect recognition of bearing images with better fractal characteristics.The experimental results show that the accuracy of the image defect recognition of bearing components obtained by the algorithm is 100%,which marks the successful recognition of the image defects of the parts.At the same time,it proves that the improved MF DFA algorithm has certain practical significance and value.

关 键 词：多重分形去趋势波动分析算法 三角覆盖 经验模态分解 滑动窗口 图像识别 

分 类 号：U693.1[交通运输工程—港口、海岸及近海工程]