基于自模板重构与NSCT的汽车内饰件表面缺陷检测方法研究

汽车内饰件作为汽车产业中重要的零部件之一,不仅要具备良好的尺寸精度和物理性能,而且要满足很高的外观需求。由于汽车内饰件表面具有高反射性且缺陷种类繁多,准确、快速地识别和定位表面缺陷具有较高的挑战性。为此提出一种基于自模板重构与非下采样Contourlet变换(NSCT)的汽车内饰件表面缺陷检测方法。利用稀疏表示算法对试件图像进行重构,生成自模板;利用差影法生成残差图像,以抑制背景信息;利用NSCT对残差图像进行增强,实现缺陷的分割与定位。该方法对图像的光照不均不敏感,且在残差图像生成过程中无需对图像进行位置较准,因此,该方法具有较强的鲁棒性。实验结果表明,该方法可以有效地检测汽车内饰件表面的各种缺陷,其检测精度可达0.1 mm。

SURFACE DEFECT DETECTION OF AUTOMOBILE INTERIOR PARTS BASED ON SELF-TEMPLATE RECONSTRUCTION AND NSCT

As one of the most important parts in automobile industry, automobile interior parts not only have good dimensional accuracy and physical properties, but also meet high appearance requirements. Because the surface of automobile interior parts has high reflectivity and many kinds of defects, how to identify and locate surface defects accurately and quickly is a challenge. This paper presented a surface defect detection of automobile interior parts based on self-template reconstruction and NSCT. We used sparse representation algorithm to reconstruct the sample image, generate self-template, and used image subtraction algorithm to generate residual image to suppress background information. Then the residual image was enhanced by NSCT, and the defect was segmented and located. The method is insensitive to the uneven illumination of the image, and does not need to be accurate in the process of residual image generation, so it has strong robustness. The experimental results show that the method can effectively detect various defects on the surface of automobile interior parts, and its detection accuracy can reach 0.1 mm.