Compressive thermal wave imaging for subsurface analysis

Abstract

Subsurface detail extraction in active thermography demands high capturing rates, resulting in less exposure time, data redundancy, large bandwidth requirement, and wastage of sensing resources and memory. Compressive sensing (CS) is a data acquisition technique that overcomes these limitations by acquiring the signal at sub-Nyquist rates with fewer measurements considering the signal is sparse in some transformed domains and reconstructing the original response. This paper validates the application of CS in frequency modulated thermal wave imaging by experimenting on a quick responsive mild steel specimen with artificially simulated back hole defects. The discrete cosine transform is selected as sparsity prior and the orthogonal matching pursuit is used to reconstruct the original thermal response from the compressed measurements. The initial analysis is carried out on choosing the optimal sparsity parameter and the number of measurements, followed by the effect of the number of measurements on defect detection in various post-processing techniques used in frequency modulated stimulus. The defects detected and their qualitative analysis through assessing signal-to-noise ratio confirmed the suitability of CS for efficient reconstruction of thermal data and thereby enhancing the defect signature.