ABSTRACT In order to effectively conduct the defect detection of nuclear fuel pellets end face and avoid the leakage of nuclear radiation, a defect detection system for the nuclear fuel pellets end face based on machine vision is proposed. Firstly, aiming at the complexity of the defect detection of nuclear fuel pellets, a set of image acquisition system lighted by left-right symmetric grating is designed. Then, after fusing the images of left-right structured light those cross points are extracted which classified based on the Gaussian mixture model (GMM). Finally, a series of morphological operations such as dilation operation are conducted for the classified points to obtain the defect area of nuclear fuel pellets end face. The experimental results show that this method reduces the influence of complex characteristics of form, texture, and color of the sample end face on the defect detection and relatively good detection results are gained for various defects with 99.5% accuracy. It takes less than 0.4 s to fully meet the requirements of industrial automation testing.

中文翻译：

---

基于机器视觉的核燃料芯块端面缺陷检测系统

摘要 为了有效地进行核燃料芯块端面缺陷检测，避免核辐射泄漏，提出了一种基于机器视觉的核燃料芯块端面缺陷检测系统。首先，针对核燃料芯块缺陷检测的复杂性，设计了一套左右对称光栅照明的图像采集系统。然后，将左右结构光图像融合后，提取交叉点，并基于高斯混合模型（GMM）进行分类。最后，对分类的点进行膨胀运算等一系列形态运算，得到核燃料芯块端面的缺陷面积。实验结果表明，该方法降低了形态、纹理、缺陷检测上样品端面颜色和颜色，对各种缺陷获得了较好的检测结果，准确率为99.5%。不到0.4s即可完全满足工业自动化测试要求。