基于AHSVM-WOA的钢板表面缺陷分类算法

Classification algorithm based on AHSVM-WOA for steel plate surface defects

为了降低标签和特征噪声对钢板表面缺陷分类的影响,提出一种抗噪声的超球体支持向量机(anti-noise hypersphere support vector machine,简称AHSVM)分类模型.鉴于鲸鱼优化算法(whale optimization algorithm,简称WOA)能对AHSVM分类模型的参数进行寻优且能提高运行效率,提出AHSVM与WOA结合的AHSVM-WOA算法.4种分类算法对6类热轧钢板表面缺陷的分类结果表明,AHSVM-WOA算法有良好的分类效果,在抑制标签和特征噪声方面性能优良,缩短了参数选择的时间.

In order to reduce the adverse effects of label noise and feature noise in the process of classification for steel surface defects,an anti-noise hypersphere support vector machine(AHSVM)was proposed.Because whale optimization algorithm(WOA)could find the best parameters of the AHSVM classification model and improve the execution efficiency,it was combined with AHSVM to build the AHSVM-WOA algorithm.Four classification algorithms were used to classify six types of hot-rolled steel surface defects.The experiment results showed that the AHSVM-WOA classification model had the good classification performance,especially in restraining the adverse effects of label noise and feature noise.Moreover,the AHSVM-WOA algorithm reduced the time of parameter selection.