实时带钢表面质量检测算法研究与应用

Study and Application of Real-Time Steel Surface Quality Detection Algorithm

针对现有带钢表面质量检测技术在检测精度、数据实时吞吐量等方面存在的问题,提出了一种基于纹理特征编码的带钢表面缺陷检测方法。首先,利用高斯滤波器的差分响应模拟人类视觉的纹理感知模型,在原始图像空间内进行缺陷可疑点检测;然后对疑似缺陷位置进行纹理特征的提取与编码,在纹理特征编码空间内完成缺陷的精密定位。实验结果表明,该方法可以对带钢表面常见的氧化、孔洞、边裂、麻点等十几种不同类型缺陷进行精确地检测,降低了系统的误检率,同时提高了缺陷检测的识别率和算法的计算效率。

Visual sensing has become an important technique for surface defect detection of various productions including steel. Allowing for the problems of current steel surface defect detection in detection precision and efficiency of algorithm,a new method of steel surface defect detection is proposed based on coding texture features. Firstly,the texture perception model imitating human vison is built using the differential response of gaussian filter in original image space, and suspected defect regions are detected;secondly, the feature coding is accomplished in the suspected defect regions;lastly,the threshold based on mahalanobis distance is realized and the regions of interest are combined, and the detection is accomplished from coarse to fine. The experiments show that the proposed method could detect different kinds of defects including hole, oxidate, edge crack and dash, etc. The precision of detection is enhanced while the false detection rate is degraded and the computational efficiency is enhanced greatly.