Accurate thickness characterization of materials via wavelet denoising and baseline correction at the terahertz frequency

This study presents a non-destructive testing method for material thickness characterization, which combines wavelet denoising with baseline correction to enhance thickness extraction accuracy under low signal-to-noise ratio conditions. The proposed dual-stage processing framework first uses wavelet thresholding to suppress high-frequency noise. Then, adaptive iterative re-weighted penalized least squares baseline correction in the time domain eliminates low-frequency drifts and recovers weak front-surface echoes. Without baseline correction, traditional methods require peak searching within predefined local time windows. This increases algorithmic complexity and reduces accuracy when processing large datasets. Experimental validation was conducted on the thickness measurement of polyolefin rubber. Through systematic evaluation, optimal parameters were determined. Compared to the traditional time-of-flight method, the proposed approach increases the success rate of thickness characterization from 69.98% to 99.24%. The method supports full-range global peak detection, reducing complexity and improving robustness for large-scale data. This method shows significant potential for practical applications such as coating and rubber inspections, offering efficient and accurate technical support for quality control in industrial production.