基于模糊松弛约束的外观缺陷多核学习技术

为了满足当前工业生产中对产品外观缺陷检测的精度及实时的要求，提出一种基于模糊松弛约束多核学习的产品外观缺陷检测方法（FRC-MKL）。该方法对各类外观缺陷的样本图像进行特征提取，采用模糊约束理论求解组合核函数中各核函数的权重，通过给每个核函数划定一个模糊权重得到组合核函数，实现了将该组合核函数作为分类器的核函数进行缺陷的学习分类。实验结果表明，利用模糊松弛约束和多核技术可以提高产品外观缺陷的检测精度并且在实地检测中可以更好地满足外观缺陷的实时检测。

Appearance Defects Multiple Kernel Learning Technology Based on Fuzzy Relaxation Constraints

In order to meet the accuracy and real-time requirements of product appearance defect detection in current industrial production,a product appearance defect detection method based on fuzzy relaxation constrained multiple kernel learning (FRC-MKL) was proposed.This method extracts features from sample images of various appearance defects,and uses fuzzy constraint theory to solved the weights of each kernel function in the combined kernel function.By assigning a fuzzy weight to each kernel function,the combined kernel function is used as the kernel function of classifier to classify the defects.The experimental results show that the application of fuzzy relaxation constraint and multiple kernel technologies can improve the detection accuracy of product appearance defects and better satisfy the real-time detection of appearance defects in the field.