版图分类法在带钢表面缺陷识别中的应用研究  

Research on Territory Classification Method for Strip Surface Defect Recognition

作　　者：王永慧[1] 颜云辉[1] 赵大兴[2] 赵久梁[1] 

机构地区：[1]东北大学,沈阳110004 [2]湖北工业大学,武汉430068

出　　处：《中国机械工程》2010年第5期562-566,共5页China Mechanical Engineering

基　　金：国家863高技术研究发展计划资助项目(2008AA04Z135);国家自然科学基金委员会和上海宝钢集团公司联合资助项目(50574019)

摘　　要：针对现有带钢表面缺陷模式识别分类器所用算法的不足,提出了一种新型的分类算法——版图法。该算法结构简单,在时间消耗较少的情况下,具有更高的识别精度和更好的适应性。类别众多、形态复杂、类别边界交叠的带钢表面缺陷图像分类实验所得结果令人满意。In response to the dilemma for pattern recognition by single classifier toward strip surface defects,a novel classification algorithm territory method was proposed.This method is simple,less time consuming,higher recognition accuracy and adaptability.The experimental study on classifying strip surface defects was presented,and the results show that the defect classes are with complex features,even the overlapping class border.The classification results are satisfactory.

关 键 词：版图 分类 带钢 表面缺陷 识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]