Total Variation Regularized Sparse Tensor Decomposition for Eddy Current Pulsed Thermography Sequence Processing

Eddy current pulsed thermography (ECPT) is widely used in the nondestructive testing of metal surface defects. Since the defect information is sometimes disturbed by the background noise, it is necessary to separate the defect information from the background in the ECPT image sequence. Tensor robust principal component analysis (TRPCA), as a sparse tensor decomposition method, is widely used in ECPT sequence processing, but sometimes the separated defect information is too sparse and incomplete, or the high temperature area formed by heat conduction around the defect is also separated as a sparse part. Based on the sparsity and spatio-temporal continuity of defect information in ECPT image sequence, a 3D total variation regularized sparse tensor decomposition (TVRSTD) method is proposed in this article. To verify the performance of TVRSTD, multiple experiments are performed on samples containing different defects, and the proposed method is compared with some existing sparse tensor decomposition methods. Experimental results show that the proposed TVRSTD can effectively separate the defect information, and has better performance than the existing methods in terms of defect contrast and signal-to-noise ratio.