改进型序贯算法在在线表面瑕疵检测中的应用

Online Surface Defects Inspection System Based on Improved Sequential Algorithm

为了实现对高速在线卷板材表面瑕疵的检测,建立了瑕疵检测系统。该系统对图像处理部分的shading算法,序贯瑕疵标记算法等进行了改进。首先,依据实际工业环境介绍了在线卷板材瑕疵的特点,检测系统主要由相机单元、光源、显示器、编码器、图像处理器和标签机组成。然后着重介绍了每个部件的作用,和相互链接关系以及应用在图像处理器中软件部分的算法。阐述了其中shading算法的基本步骤,分析了原始序贯算法等价表的缺陷,针对缺陷提出了完备化的改进方案,并给出了完备化方案的证明。实验结果表明:完备后的算法可以准确标记复杂图像瑕疵,给出正确的标记结果以及瑕疵的相关信息。完备后的算法经过理论和实验证明是正确的,经试验平台的验证该系统可以应用在高速在线卷板材瑕疵检测上。

In order to realize the online high-speed plate surface defects detection, a defects detection system is established and the system improves the algorithms of the image processing, such as shading algorithm, sequential algorithms. First, based on the real industrial environment, the defect characteristics of the plate on the spool and the composition of the detection system are presented, the test system is mainly composed of camera unit, light unit, display unit, encoder unit, image processor unit, and label making machine. The role of each components, link relations and the algorithm which is used in the software part of image processor are introduced. The basic steps of shading algorithm are expounded. And the equivalence table fault of the original sequential algorithm is an- alyzed. In view of the fault, a self-contained scheme is put forwarded, the proof of completion scheme is also given. The experimental results show that： the algorithm after completed can accurately label the defects of complex image, present the correct labeled results as well as the related information of defects. The self-contained scheme is proved to be correct through theory and experiment, and the defects detection system can be used on high-speed on-line detection.