Abstract
Reverberation suppression is necessarily required for detecting the underwater low-speed small targets (e.g., frogmen and UUVs). In recent years, the promising reverberation suppression method is to use the correlation characteristics between multi-ping echoes to implement the low-rank sparse matrix decomposition. However, almost no prior knowledge about the moving continuity characteristic of the target has been considered in existing methods, which leads to performance degradation in complicated underwater scenarios. To solve the problem, we introduce the detecting contiguous outliers in the low-rank representation (DECOLOR) to improve the calculation accuracy of the low-rank and sparse matrix decomposition. In DECOLOR, a priori knowledge that the slow small target is moving contiguously with relatively a small size (i.e., a highlight point moving continuously across a group of sonar images), is adopted. And the locations of target highlight point in the group of sonar images are modeled by the first-order Markov Random Fields (MRFs). By doing so, the low-rank and sparse matrix decomposition using the DECOLOR shows a higher accuracy for matrix decomposition. Thus, the proposed method shows a better reverberation suppression ability than existing methods. The proposed methods are tested and evaluated through a series of simulation experiments.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have