Full Physical Model Based Data Processing For Infrared Thermography

In the field of nondestructive testing,pulsed eddy current thermography detection technology is a new technology.It has many advantages such as non-contact,high efficiency and large single detection range.It has attracted extensive attention of researchers at home and abroad,and many data processing methods based on this technology have been proposed.From the first manual selection of images to the frequency domain analysis based on Fourier transform to the widely used PCA,ICA and other methods,although they can achieve rapid defect detection,the reliability of defect information extracted by the above mathematical or statistical model detection methods is weak.Therefore,it has great significance to establish a more reliable phsical model for rapid defect detection.This paper studies and analyses the framework of fast defect detection algorithm based on full physical model systematically by means of defect feature analysis,algorithm model establishment and experimental verification.Various data processing methods are used for defect detection of different types of specimens.It is proved that the proposed method can realize rapid defect detection and measurement.The main contents are as follows:(1)Research on theoretical basis and content.Firstly,the detection principle of pulsed eddy current thermography technology is studied,and its main components and parameters are elaborated based on the experimental platform of pulsed eddy current thermal imaging.Secondly,the data characteristics commonly existed in thermal image sequences are described in detail from the spatial and temporal perspectives,which lays a theoretical foundation for the subsequent use of cascade classifiers to select features for defect recognition.Definition and mechanism of thermal emissivity were defined.The influence of uneven thermal emissivity on defect detection results was analyzed with experimental data.(2)Research on thermal image processing algorithm model based on full physical model.Firstly,the theoretical research of cascade classifier algorithm is studied to understand its application in other scenarios;then,combining with the characteristics of defect data,the algorithm framework of cascade classifier is analyzed,and the basis of feature selection is elaborated.Finally,the method proposed in this paper,i.e.the infrared thermal image data processing algorithm based on full physical model,is emphatically studied.(3)The experimental results of the proposed algorithm are presented.Pulsed eddy current thermography experiments were carried out on four kinds of artificial defect specimens with different defect types,and the corresponding defect endpoint information was extracted according to the algorithm proposed in this paper.Secondly,the sensitivity was analyzed by constantly adjusting the threshold parameters,and the results showed that the time correlation coefficient threshold is more sensitive.Finally,the detection results of the algorithm and the commonly used detection algorithms were evaluated by using three image evaluation indicators: mean square error,peak signal-to-noise ratio and information entropy,which proved that the method proposed in this paper had the best detection effect.