Improved Adaptive Finch Clustering Sonar Segmentation Algorithm Based on Data Distribution and Posterior Probability

Round 1

Reviewer 1 Report

When reading the Article “Improved Adaptive Finch Clustering Sonar Segmentation Al-2 gorithm Based on Data Distribution and Posterior Probability”, dedicated to improving the recognition of underwater objects using sonar, a number of questions arise:

1. It is not entirely clear who is the author of the circuit in Figure1. There is a link (line 127) to article number 29-31. But these are three different articles by different authors.

2. The same applies to Figure.2.

3. Also, there is no description of the operation of the algorithm shown in fig. 2. In particular, it is not clear when Process 1 is turned on in the algorithm, and when Process 2 is turned on. Does this matter and what is the difference between the method proposed by the authors and that described in Figure1 and Figure.2?

4. The caption to Table 3 clearly looks like a recommendation from the editors of the journal, and not from the authors of the article: “Table 3. This is a table. Tables should be placed in the main text near to the first time they are cited." (line 544).

5. It is not clear what the authors wanted to prove in Table 3, since they presented the data of the MRF method (from [24]), which, both in terms of (MIoU) and (Detecting speed), significantly exceed the results obtained by the method proposed by the authors.

Probably, the method proposed by the authors, based on changing the gray background border, has the right to exist in some specific conditions. For example, in conditions of limited power or speed of the used computing resource. But I would like to know more about this, for example, if the authors will enter this key parameter into the result tables.

Also, it should be noted that the authors at the beginning of the article announce:

1. A gray scale correction and data distribution calculation method based on side-scan sonar image is proposed. (line 87).

2. An adaptive Finch clustering rough segmentation algorithm based on data distribution is proposed. (line 91).

3. An improved Finch clustering target detection and target discrimination method based on a posteriori probability of side-scan sonar images is proposed. (line 96).

But only one thing is stated in the conclusions: “In this paper, an improved adaptive Finch clustering target detection algorithm 579 based on data distribution and posterior probability is proposed for CPU and low perfor-580 mance GPU” (line 579), and further it is  explained how this has been achieved.

Naturally, the question arises, what did the authors mean, the first three achievements or the last one?

Also, the authors in lines 25-26 write: “…and the detection accuracy reaches thedomestic advanced level, and meets the real-time detection requirement.” For a report to the employer, this is normal. But for a scientific article, one should not write about regional achievements. Only universal human achievements, in no way limited by the place of their receipt, are of value in science. Although, I repeat, this is quite acceptable for technical achievements in the industry, but not in science.

In very interesting experimental results, the authors present data on the parameters of using the Shark-S900U side-scan sonar used in their experiment. “The Shark-S900U side scan sonar adopted in this experiment is set at a fixed height of 8m, a scanning Range of 30m and a scanning frequency of 900kHz. The 168 110 ping data received by 6s is spliced into 6s narrowband image with a resolution of 169 4800*110". However, the authors leave aside the accuracy with which this experiment should be carried out. It is not clear what will happen if you increase the height from height of 8m, to 10 m, or reduce it to 6m? The same applies to the other parameters of the experiment. What is its reproducibility range? Maybe to get the desired result, you should change the size of 4800*110 to 9600*55 or 2400*220?

We are talking about the fact that the authors analyze the operation of a particular device, not paying attention to the physical processes that limit its resolution. General mathematical methods of signal processing give the authors a positive result, but this result cannot be considered the best possible. Whereas the analysis of the physical features of acoustic signals at the general physical level can raise the authors' research to a new height. However, this last remark refers more to wishes and does not negate the results obtained by the authors.

In general, we can say that the algorithm proposed by the authors can facilitate the data processing process when using low-performance calculators, and the Article "Improved Adaptive Finch Clustering Sonar Segmentation Al-2 gorithm Based on Data Distribution and Posterior Probability" can be published in the Electronics journal after revision the authors of the submitted text.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The manuscript presents an improved adaptive Finch clustering target detection algorithm based on data distribution and posterior probability. Some suggestions are shown as follows.

1.  In Section I, the authors are suggested to identify the main challenge of segmentation  between the common image and the sonar image.

2. In page 2, the authors summarize several related works, which are suggested to briefly present their connections with the proposed work. 

3. What is the relationship between process 1 and 2 as shown in Figure 2.

4. Eq.(1) and other equations are suggested to be explained by all symbols and notation's physical meanings.

5. Figure 5 should be not separated and the title should be in English.

6. How to choose the covariance matrix S as shown in Eq. (3).

7. In Section 3.3.1, the authors present the posterior probability, which is suggested to briefly summarize to highlight the connection with the proposed algorithm and the fundamental and straightforward equations can be ignored.

8. In the experiment part, the contribution of GPU embedded on NVIDIA Jetson TX2 is suggested to highlight. 

The writing of the manuscript should be carefully revised in terms of wording, grammar, and structure.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf