基于热图时序特征和PNN的孔洞缺陷红外无损检测方法

利用热图时序特征和PNN，提出了一种以像素为单位，实现缺陷红外无损检测的新方法。该方法首先采用红外热像仪获取加热试件在降温过程中的红外时序热图；其次，提取时序热图中正常和异常区域的灰度值，建立不同区域的灰度值与时间的关系，进而获得相应的初始特征；再次，采用主成分分析方法对初始特征进行提取，获得时序特征；最后，以时序特征作为训练样本，构建概率神经网络，实现孔洞缺陷检测。实验结果表明，正常区和异常区识别率分别可达到95%和85%。

Infrared Nondestructive Testing for Hole Defect Based on Temporal Characteristics and Probabilistic Neural Networks

This paper presents a novel method of infrared NDT for detecting hole defects based on temporal charac-teristics and probabilistic neural network （PNN）. Firstly, the sequence image was obtained by thermal imaging camera. Secondly, the gray value of normal and abnormal area was extracted and different parts of the gray value of time were set up, and then the initial characteristics were achieved. The principal component analysis was used to extract initial characteristics and get the temporal characteristics. Finally, the temporal characteristics were adopt-ed as the training sample, and the probabilistic neural network was founded for the hole defect detection. Results showed that the recognition rates of the normal and abnormal area were 95%and 85%respectively.