【作者】 徐科； 杨朝霖； 周鹏；

【Author】 Xu Ke;Yang Chaolin;Zhou Peng;National Engineering Research Center for Advanced Rolling Technology,University of Science and Technology Beijing;

【机构】 北京科技大学高效轧制国家工程研究中心；

【摘要】 针对连铸坯表面图像背景复杂、冗余性强等特点,提出了一种基于ICA-SVM的表面裂纹在线自动算法。采用独立主元分析方法对连铸坯表面图像进行特征提取,建立特征子空间,然后选用高斯径向基核函数的支持向量机对测试样本集进行缺陷分类。对裂纹、氧化铁皮、振纹、残渣、水痕和光照不均等实验结果表明,当选取独立成分个数为30时,表面裂纹的识别率可以达到96%以上。更多还原

【Abstract】 A surface crack recognition algorithm based on ICA-SVM is proposed for the features of the surface defect images with large amount of data redundancy in continuous casting Slabs. This paper uses ICA method to extract the features of the original images and constructs the feature subspace, then selects a SVM method with gaussian radial basis function to classify the defects of test samples set. The experiment proves that the recognition rate can reach over 96% when 30 is selected as the number of independent components.更多还原