基于分类与协同表示的刀具图像超分辨率重建  

Super-Resolution Reconstruction of Tool Image Based on Classification and Collaborative Representation

作　　者：李渊 袁德志 朱锟鹏 LI Yuan;YUAN Dezhi;ZHU Kunpeng(School of Machinery and Automation,Wuhan University of Science and Technology,Wuhan 430081,China;Institute of Intelligent Machinery,Hefei Institute of Physical Science,Chinese Academy of Sciences,Hefei 230031,China)

机构地区：[1]武汉科技大学机械自动化学院,武汉430081 [2]中国科学院合肥物质科学研究院智能机械研究所,合肥230031 [3]常州先进制造技术研究所智能装备技术研究中心,常州213164

出　　处：《组合机床与自动化加工技术》2025年第2期75-80,共6页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：国家自然科学基金项目(52175528);中国科学技术大学学生创新创业和成果转化行动计划资助项目(CY2023X003)。

摘　　要：获取高分辨率的微细铣刀图像是使用视觉方法精确监测刀具磨损状态的关键,在微铣削加工过程中,由于主轴转速大、刀具直径细小,高分辨率的刀具图像是难以获取到的。为了解决此问题,提出了一种基于分类与协同表示的刀具图像超分辨率重建算法。根据刀具图像中的图像块存在多样性,对其进行分类并训练得到相对应的字典,解决了单一字典对特征表示不足的缺点。使用协同表示的方法求得每一类字典的映射矩阵,加快重建速度。最后,考虑到刀具图像中存在许多重复的结构,添加自相似约束来提高重建效果。实验结果表明,与其它传统算法相比,提出的算法不仅有着最高的峰值信噪比和结构相似性指标,同时也有着更好的刀具边缘视觉效果。Obtaining high-resolution micro-endmill images is crucial for accurately monitoring the tool wear status using visual methods.In the micro milling process,it is difficult to obtain high-resolution tool images due to the high spindle speed and small tool diameter.To address this issue,a tool image super-resolution reconstruction algorithm based on classification and collaborative representation is proposed.Considering the diversity of image patches in tool images,they are classified and corresponding dictionaries are trained to overcome the limitation of a single dictionary in feature representation.The mapping matrices for each class of dictionaries are obtained using collaborative representation,which accelerates the reconstruction process.Furthermore,to enhance the reconstruction effect,self-similarity constraints are added,considering the presence of many repetitive structures in tool images.Experimental results demonstrate that compared to other traditional algorithms,the proposed algorithm not only achieves the highest peak signal-to-noise ratio and structural similarity index but also exhibits better visual effects on tool edges.

关 键 词：微铣削 超分辨率 分类 协同表示 

分 类 号：TH16[机械工程—机械制造及自动化] TG71[金属学及工艺—刀具与模具]