基于SVD算法的红外热波无损检测方法研究

Research of infrared thermal wave nondestructive testing method based on SVD algorithm

针对采用红外热波无损检测技术对材料缺陷进行特征提取的技术空白,提出了一种新的基于奇异值分解(SVD)的红外序列图像特征提取方法。研究表明对重构的缺陷相空间矩阵进行奇异值分解,其空间与时间基向量包含了缺陷静态空间与动态热量变化的特征信息。在缺陷代数特征的基础上,提取具有时空信息的特征值构造缺陷特征向量。实验分析表明,通过对热障涂层缺陷进行特征提取,在运用RBF神经网络进行缺陷的分类验证中取得了较好的效果。

Concerning the gap of feature extraction for material defect by adopting infrared thermal wave nondestructive testing technology,a new method of feature extraction for infrared image sequence is put forward on the basis of singular value decomposition（SVD）.The research shows that characteristic vector reflects the information of the defects static space and dynamic heat change.On the basis of algebra features of the defects,tectonic feature vector of defects with space-time characteristics is extracted.The analyzing results of the experiment indicate that the application of this method in RBF neural network is verified to be better in effect through feature extracting thermal barrier coating defect type.