Identification of Surface Defect for Steel Strip via Multi-feature Set and MK-SVM
-
摘要: 针对带钢表面缺陷难以有效辨识的问题,设计基于混合域特征集与多核支持向量机的带钢表面缺陷分类辨识方法。首先基于灰度特征、纹理特征计算带钢表面缺陷的图像特征指标量,构造混合域特征集,再将混合域特征集输入给多核支持向量机实现带钢表面缺陷的分类辨识。实验结果表明,该方法能够有效提取带钢表面缺陷的低维敏感特征,辨识精度高,泛化能力强,可以应用于工程企业带钢表面缺陷的分类辨识。Abstract: In order to solve the problem that the surface defects for steel strip are difficult to identify effectively, based on mixed domain feature set and multi-kernel support vector machine (MK-SVM), a classification and identification method for surface defects of steel strip was proposed. Firstly, the image feature indexes of surface defects for steel strip were calculated based on gray level features and texture features, and the mixed domain feature set was constructed, and then the mixed domain feature set was input to multi-kernel support vector machine to achieve the classification and identification of surface defects for steel strip. The experimental results show that the method can effectively extract the low dimensional sensitive features of surface defects for steel strip, with high identification accuracy and strong generalization ability, and can be applied to the classification and identification of surface defects steel strip in engineering.
-
表 1 带钢表面的缺陷类型及特征描述
缺陷类型 特征描述 划痕 在形态上,其宽度比较均匀、多为平行于轧制方向的浅痕。 补丁 补丁在钢板表面分布无特定规律,形状也比较多样,缺陷面积较大。 夹杂物 一般为非金属夹杂物,其形貌多样,常沿轧制方向排列。 麻点 麻点沿带钢纵向连续成片分布,单个缺陷面积较大,形成初期呈凹坑状,在后期脱落后在带钢表面会留下凹坑。 银纹 通常表现为一系列长短不同、深浅不一的细小裂口,一般呈水波纹或鱼鳞状。 轧制鳞片 该缺陷形貌多样,主要为鱼鳞状,还有部分呈方块状、条形状。 表 2 带钢表面不同缺陷图像的灰度特征参数
缺陷类型 平均值 方差 倾斜度 峭度 能量 熵 划痕 70.702 970.160 3.0281 12.2925 0.0298 5.6890 补丁 145.631 3452.341 0.3376 1.8785 0.0080 7.4572 夹杂物 73.140 92.934 0.2584 2.0949 0.0287 5.2754 麻点 214.731 785.248 −0.3134 2.0949 0.0135 6.5417 银纹 161.577 800.208 0.2285 2.9910 0.0100 6.8486 轧制鳞片 118.677 238.850 −0.0921 2.9928 0.0182 5.9884 表 3 带钢表面不同缺陷的纹理特征参数
缺陷类型 2阶矩 对比度 相关度 纹理熵 纹理方差 逆差矩 划痕 0.3629 0.2954 0.2498 1.7656 27.701 0.9099 补丁 0.0256 1.7315 0.0708 4.0453 104.350 0.6177 夹杂物 0.3484 0.0990 2.0231 1.3565 25.828 0.9505 麻点 0.1130 0.2290 0.3251 2.4813 194.814 0.8928 银纹 0.0384 1.6364 0.2339 3.5672 114.698 0.5917 轧制鳞片 0.1275 0.4558 0.7637 2.3827 63.179 0.7930 表 4 基于混合域特征集的特征向量.
缺陷类型 归一化特征向量 平均值 方差 能量 熵 2阶矩 对比度 相关度 纹理熵 纹理方差 逆差矩 划痕 0.2059 0.2576 0.6097 0.3662 0.6807 0.1203 0.1128 0.2607 0.1067 0.4614 补丁 0.4242 0.9167 0.1637 0.4800 0.0480 0.7049 0.0320 0.5972 0.4019 0.3132 夹杂物 0.2130 0.0247 0.5872 0.3396 0.6535 0.0403 0.9136 0.2003 0.0995 0.4820 麻点 0.6255 0.2085 0.2762 0.4211 0.2120 0.0932 0.1468 0.3663 0.7503 0.4527 银纹 0.4707 0.2125 0.2046 0.4408 0.0720 0.6662 0.1056 0.5266 0.4417 0.3000 轧制鳞片 0.3457 0.0634 0.3723 0.3855 0.2392 0.1856 0.3449 0.3518 0.2433 0.4021 表 5 3种SVM对各类缺陷的正确分类量
缺陷
类型分类精度/% 多项式SVM 径向基SVM 多核SVM 划痕 64 71 91 补丁 94 96 96 夹杂物 98 97 97 麻点 96 93 88 银纹 95 96 99 轧制鳞片 100 100 100 -
[1] MENTOURI Z, MOUSSAOUI A, BOUDJEHEM D, et al. Steel strip surface defect identification using multiresolution binarized image features[J]. Journal of Failure Analysis and Prevention, 2020, 20(6): 1917-1927. doi: 10.1007/s11668-020-01012-7 [2] FENG X L, LUO L, GAO X W. SDDA: a method of surface defect data augmentation of hot-rolled strip steel[J]. Journal of Electronic Imaging, 2022, 31(3): 033002. [3] SUN X H, GU J N, TANG S X, et al. Research progress of visual inspection technology of steel products-A review[J]. Applied Sciences, 2018, 8(11): 2195. doi: 10.3390/app8112195 [4] CHEN Y J, DING Y Y, ZHAO F, et al. Surface defect detection methods for industrial products: a review[J]. Applied Sciences, 2021, 11(16): 7657. doi: 10.3390/app11167657 [5] WANG W Y, LU K, WU Z H, et al. Surface defects classification of hot rolled strip based on improved convolutional neural network[J]. ISIJ International, 2021, 61(5): 1579-1583. doi: 10.2355/isijinternational.ISIJINT-2020-451 [6] 赵朗月, 吴一全. 基于机器视觉的表面缺陷检测方法研究进展[J]. 仪器仪表学报, 2022, 43(1): 198-219.ZHAO L Y, WU Y Q. Research progress of surface defect detection methods based on machine vision[J]. Chinese Journal of Scientific Instrument, 2022, 43(1): 198-219. (in Chinese) [7] KONOVALENKO I, MARUSCHAK P, KOZBUR H, et al. Defectoscopic and geometric features of defects that occur in sheet metal and their description based on statistical analysis[J]. Metals, 2021, 11(11): 1851. doi: 10.3390/met11111851 [8] TSAI D M, FAN S K S, CHOU Y H. Auto-annotated deep segmentation for surface defect detection[J]. IEEE Transactions on Instrumentation and Measurement, 2021, 70: 5011410. [9] BAKAY M S, AĞBULUT Ü. Electricity production based forecasting of greenhouse gas emissions in Turkey with deep learning, support vector machine and artificial neural network algorithms[J]. Journal of Cleaner Production, 2021, 285: 125324. doi: 10.1016/j.jclepro.2020.125324 [10] HAN T, ZHANG L W, YIN Z J, et al. Rolling bearing fault diagnosis with combined convolutional neural networks and support vector machine[J]. Measurement, 2021, 177: 109022. doi: 10.1016/j.measurement.2021.109022 [11] 陈法法, 杨晶晶, 肖文荣, 等. Adaboost_SVM集成模型的滚动轴承早期故障诊断[J]. 机械科学与技术, 2018, 37(2): 237-243. doi: 10.13433/j.cnki.1003-8728.2018.0212CHEN F F, YANG J J, XIAO W R, et al. Early fault diagnosis of rolling bearing based on ensemble model of Adaboost SVM[J]. Mechanical Science and Technology for Aerospace Engineering, 2018, 37(2): 237-243. (in Chinese) doi: 10.13433/j.cnki.1003-8728.2018.0212 [12] 刘涛, 沈朝洋, 雷经发, 等. 碳钢试样拉伸过程磁信号多尺度熵特征提取及磁畴观测[J]. 机械科学与技术, 2022, 41(5): 808-814.LIU T, SHEN C Y, LEI J F, et al. Multi-scale entropy feature extraction of magnetic signal and observation of magnetic domain in tension of carbon steel specimen[J]. Mechanical Science and Technology for Aerospace Engineering, 2022, 41(5): 808-814. (in Chinese) [13] 李聪, 马勋, 杨锐, 等. 基于形状和纹理特征的正庚烷环形油池火特性[J]. 清华大学学报(自然科学版), 2021, 61(6): 502-508. doi: 10.16511/j.cnki.qhdxxb.2020.22.042LI C, MA X, YANG R, et al. Characterization of n-heptane annular pool fires based on flame shape and texture features[J]. Journal of Tsinghua University (Science and Technology), 2021, 61(6): 502-508. (in Chinese) doi: 10.16511/j.cnki.qhdxxb.2020.22.042 [14] BOUDIAF A, BENLAHMIDI S, HARRAR K, et al. Classification of surface defects on steel strip images using convolution neural network and support vector machine[J]. Journal of Failure Analysis and Prevention, 2022, 22(2): 531-541. doi: 10.1007/s11668-022-01344-6 [15] 何彧, 宋克臣, 张德富, 等. 融合多层级特征的弱监督钢板表面缺陷检测算法[J]. 东北大学学报(自然科学版), 2021, 42(5): 687-692. doi: 10.12068/j.issn.1005-3026.2021.05.012HE Y, SONG K C, ZHANG D F, et al. Weakly-supervised steel plate surface defect detection algorithm by integrating multiple level features[J]. Journal of Northeastern University (Natural Science), 2021, 42(5): 687-692. (in Chinese) doi: 10.12068/j.issn.1005-3026.2021.05.012