Abstract
During terahertz (THz) non-destructive testing (NDT), multiple echoes from the sample interface reflection signals are mixed with the detection signals, resulting in signal distortion and affecting the accuracy of the THz NDT results. Combined with the frequency property of multiple echoes, an improved wavelet multi-scale analysis is put forth in this paper to correct multiple echoes, allowing the maximum retention of detailed signal information in contrast with the existing echo correction methods. The results showed that the improved wavelet multi-scale analysis enhanced the continuity and smoothness of the image at least twice in testing adhesive layer thickness, prevented missing judgments and misjudgments in identifying characteristic defects, and ensured accurate detection results. Hence, it is of great significance for evaluating the THz NDT results.