Defect feature extraction of marine protective coatings by terahertz pulsed imaging

Tu, Wanli and Zhong, Shuncong and Incecik, Atilla and Fu, Xibin (2018) Defect feature extraction of marine protective coatings by terahertz pulsed imaging. Ocean Engineering, 155. pp. 382-391. ISSN 0029-8018 (https://doi.org/10.1016/j.oceaneng.2018.01.033)

[thumbnail of Tu-etal-OE-2018-Unmarked-defect-extraction-of-marine-protective-coatings]
Preview
Text. Filename: Tu_etal_OE_2018_Unmarked_defect_extraction_of_marine_protective_coatings.pdf
Accepted Author Manuscript
License: Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 logo

Download (1MB)| Preview

Abstract

Feature extraction of marine protective coatings using Hilbert transform (HT) and wavelet transform modulus maximum (WTMM) on terahertz pulsed imaging (TPI) of the coatings was proposed. For TPI-based marine protective coating detection, it is difficult to locate exactly the reflected echoes form internal structure of coating system by the faint feature in the time domain due to the interference of background noise. However, those faint features were usually caused by the interface between two different medium layers whose refractive indices are very similar or caused by micro defects beneath the coating. The proposed algorithm was validated by simulated and experimental TPI waveform obtained from marine paint samples with or without defects. To extract the structure feature more clearly and intuitively, Hilbert transform procedure was carried out on detected terahertz signal to get Hilbert envelope for further processing. Subsequently, the modulus maximum of the stationary wavelet transform approximation coefficients were employed as the criteria for feature extraction of internal interfaces and defect features, according to the relationship between WTMM and signal singularity. The results demonstrated that the combination of HT and WTMM algorithms could be used to exactly extract the structure feature and to evaluate the position of defects beneath coatings.