Current Issue Cover


摘 要
目的 钢板表面缺陷的种类多样、灰度结构复杂,现有的图像分割技术在钢板缺陷图像识别上存在不足,本文结合图像灰度矩阵的空间特征,提出一种基于三维灰度矩阵的钢板表面缺陷识别算法。 方法 首先根据灰度图像构建三维灰度矩阵;然后引入半类间方差改进克里金插值算法,绘制三维灰度矩阵的等值线图;接着构建等值线的拓扑关系树;最后根据自定义的全局搜索策略和局部搜索策略相结合,寻找局部凹凸区域,从而定位缺陷区域,达到分割钢板表面缺陷的目的。 结果 通过对氧化、辊印、结疤和气泡四类钢板缺陷图像进行测试,从分割效果和评价指标两方面对比其他钢板缺陷分割算法,本文方法能更有效的识别缺陷区域,对光照变化不敏感,在保证低误差率的前提下,提高了有效分割率。 结论 本文提出的基于三维灰度矩阵的钢板缺陷图像识别算法可以有效的识别多种类型的钢板缺陷,即使在缺陷结构复杂的图像识别中仍具有较高识别率。
Image Recognition of Steel Plate Defects Based on 3D Gray Matrix


Objective There are many kinds of surface defects and complicated gray structure in steel plate. The existing image segmentation technology has some shortcomings in the image recognition of steel plate defects. This paper proposes a surface defect recognition algorithm based on the spatial characteristics of 3D gray matrix. Method Firstly, a three-dimensional gray matrix is constructed according to the gray image; then a half-class variance improved Kriging interpolation algorithm is introduced to draw the contour map of the three-dimensional gray matrix; then the topological relationship tree of the contours is constructed; finally, the customized global search strategy and the local search strategy are combined to find the local concave and convex areas, thereby locating the defect area and achieving the purpose of dividing the surface defects of the steel plate. Result By testing the defect images of four types of steel plates, such as oxidation, roll printing, crusting and air bubbles, compared with other steel plate defect segmentation algorithms from the aspects of segmentation effect and evaluation indicator, this method can identify defect areas more effectively and is not sensitive to illumination changes. Under the premise of ensuring a low error rate, the effective segmentation rate is improved. Conclusion The steel plate defect image recognition algorithm based on three-dimensional gray matrix can effectively identify many types of steel plate defects, even in the image recognition with complex defect structure.