Abstract:In the process of pipeline defect detection using ultrasonic guided wave, interference noise is mixed in, which seriously affects the extraction and identification of defect reflection signal. In this paper, a new de-noising method for pipeline defect detection signal based on matching pursuit is proposed. The Hanning windowed sinusoidal excitation signal is used as the basic atom to form a self-defined over complete waveform dictionary after time translation and amplitude modulation. On this basis, the effective components are extracted by matching pursuit sparse decomposition from the pipeline defect detection signal, and the remaining noise are discarded,so as to realize signal de-noising. The finite element numerical simulation and the experimental pipeline defect detection signals are de-noised by traditional wavelet threshold and the matching pursuit method proposed in this paper respectively. The results show that the matching pursuit method is obviously better than the wavelet threshold method when the signal-to-noise ratio is low, even though the interference noise completely submerges the defect reflection echo, the matching pursuit method is still can extract the clear reflection echo of defects, it provides a new way to correctly analyze and evaluate the pipeline safety operation.