冷轧带钢表面相似线性缺陷检测

Similar Linear Defect Inspection of Cold-Rolled Strip Steel Surfaces

针对现有冷轧带钢表面的相似线状缺陷检测精度与识别率差的问题,提出一种局部二制模式LBP直方图特征与支持向量机SVM相结合的检测算法。通过对采集的大量划伤与夹杂的带钢表面缺陷图进行预处理,获得感兴趣区域,再进一步利用LBP等价模式获得目标区域的LBP直方图信息,结果显示可以很好地分辨缺陷与非缺陷,并描述的各种缺陷具有可分辨性。采用核函数为径向基函数核的SVM分类器训练识别,结果表明:该方法对划伤和夹杂的缺陷检测准确率达98%。

In order to improve the detection accuracy and recognition rate of similar linear defects on the surface of the existing cold-rolled strip steel.A detection algorithm combining the LBP characteristics of local binary pattern with the support vector machine SVM is proposed.By pre-processing a large number of scratches and mixed strip edgy surface defects,obtaining the area of interest,and further using the LBP equivalent model to obtain the LBP histogram information of the target area,the results show that the defects and non-defects can be distinguished well,and the various defects described are distinguishable.The SVM classifier training and identification of the core of the radial base function is used,and the results show that the detection accuracy of the defect slicing and inclusion is 98.