To solve the problem that the evaporation duct data is easily disturbed by noise in compressed sensing and traditional reconstruction methods have poor performance in denoising, an adaptive regularized matching pursuit denoising method is proposed and it based on the similarity threshold. This method can gradually expand the candidate set by using the adaptive idea when the signal sparsity is difficult to be known. At the same time, some atoms are removed by setting the similarity threshold, and the support set atoms are screened by the regularization process, so that the reconstruction of noise components is better constrained and the reconstruction accuracy of signal is improved. Theoretical analysis and experiments show that the proposed method has better reconstruction performance than the existing similar reconstruction methods and has better denoising performance than the wavelet denoising method. The proposed method can obtain higher reconstruction SNR under the same conditions, and can effectively realize the denoising and reconstruction of the evaporation duct data.