Research On Image Processing Algorithm Of Stainless Steel Weld Defects Detection In Eddy Current Pulsed Thermography Testing Technology

The quality of stainless steel weld in pressurized water reactor is the safeguard of the safe operation of nuclear reactor and safe transportation of nuclear fuel. In order to ensure the quality of stainless steel weld, it is necessary to find a more effective nondestructive testing technology which can be applied in site. Eddy current pulsed thermography, which is emerging recently, can realize testing rapidly. This technology can detect large area in one time, which makes it effective. Moreover, the testing result, which is figure, is more intuitive. And the image processing algorithm is very important to detect the defects in stainless steel weld by this method.This paper adopts eddy current pulsed thermography to detect stainless steel weld defect. And image processing algorithm of stainless steel weld defects is researched. The related data processing methods about preprocessing of image, extracting and enhancing feature of defect are proposed in this paper. And this paper focuses on fusing method of defect feature enhancement based on independent component analysis(ICA). The main contents are as follow:(1) The basic theory of eddy current pulsed thermography is researched. The thermal response of surface and subsurface defects with different depth are simulated and analyzed, which provide a theoretical basis for thermal image processing method. The relevant factors, which will affect the testing result, are discussed. According to the actual situation, test platform of eddy current pulsed thermography is set up by discussing the influencing factors of testing result.(2) Then, in order to improve the detection effect of defect and reduce the unnecessary interference, thermal image preprocessing methods are discussed and researched in this paper. Denoising methods of thermal image sequence are discussed in two aspects which are spatial domain and time domain. And two heat balance states based and normalization method, which are used to inhibit the influence of the surface thermal emissivity, are also discussed. And the above methods realize the denoising and inhibition of nonuniform thermal emissivity.(3) ICA method is applied to extract the information of stainless steel weld defects for thermal images which are preprocessed. A kurtosis method, which is used to identify defects automatically, is proposed. And the standard of defects identification is provided. Finally, defect feature enhancement methods based on ICA are researched mainly. Enhancement of stainless steel weld defect characteristic methods, which are wavelet fusion, Contourlet fusion and fuzzy fusion based on ICA, are proposed. For the two surface defects and subsurface defects which are applied in this paper, the wavelet fusion is more suitable for the surface defects, Contourlet fusion is more appropriate for the subsurface defects, and fuzzy fusion can enhance the feature of the four specimens. The fuzzy fusion is more applicable and universal.