Dictionary Learning-Cooperated Matrix Decomposition for Hyperspectral Target Detection

Round 1

Reviewer 1 Report

A decomposition-based detector such as dictionary learning-cooperated matrix decomposition (DLcMD) is proposed to detect hyperspectral target. Authors followed two folds where at first they exploited the low rank and sparce matrix decomposition (LRaSMD) to separate the target from background. During LRaSMD, the target atoms are updated to alleviate the impact of spectral variability.

The presented topic is interested and has application, however, I have certain questions and comments that will enhance the quality of the paper.

Firstly, the abstract is written very poor, there are many long acronyms and reader lose the actual theme of the sentence. Authors need to resolve this issue.

The manuscript widely lacks visual description of the method. I suggest the authors to add generic figures in the introduction and describe the most recent real-world challenging in hyperspectral target detection. Similarly, it is unclear for reader that how the proposed method works and the intermediate steps are not clearly highlighted. I recommend to add a detailed framework of the proposed method where each step of the method is clearly given and shows the input data, its processing from different phase, intermediate steps, and finally, the detection of the target in the input test image/video.

Methods such as 10.1109/TII.2021.3116377 and https://doi.org/10.1002/int.22537 can explore in the literature section in terms of their analysis/working made in the detection of object in visual scene.

Next, can the author add the sources of the datasets used to reach the finding of this work?

Can authors provide visual representations and description of LRaSMD model and Dictionary learning-cooperated matrix decomposition?

Finally, I found authors have added a lot of equations which seem they are taken from the existing literature, there add their supportive reference.

Author Response

Please see the attachment file.

Author Response File: Author Response.pdf

Reviewer 2 Report

Very nice new algorithm.   Very interesting use of the matrix decomposition.   the development of the GLRT algorithm was very good.

 

 

Excellent job checking out algorithm on 5 datasets.   Extensive work.

Author Response

Please see the attachment file.

Author Response File: Author Response.pdf

Reviewer 3 Report

In this paper, the authors propose a matrix decomposition-based detector, called dictionary learning-cooperated matrix decomposition for hyperspectral target detection.

The investigated problem is of great relevance for the GRS scientific community. I think that the paper is very well written and organized. The proposed mathematics seem correct. Conducted tests and obtained results are very convincing. I highly recommend the publication of the paper after two modfications:

1- The authors are requested to add some quantitative results (improvments as compared to tested literature methods) in the Abstract and Conclusion sections: this may only be beneficial for the manuscript.
2- The authors are requested to cite (and, if possible, compare to) the following related works: 10.1109/TGRS.2020.2972289 and 10.3390/rs11182164.

Author Response

Please see the attachment file.

Author Response File: Author Response.pdf