Opium Poppy Detection Using Deep Learning

Round  1

Reviewer 1 Report

In this paper a new method has been proposed for detecting the opium poppy in remote sensing images. The proposed method is based on deep learning (especially on SSD model). Although the paper presents some interesting work, it has some drawbacks as listed below:

Major reviews :

1) In the conclusion, the authors say : "Compared to existing monitoring methods, our work has three unique points:" Personally, I don't find  a comparison with existing method

2) In relation with the first major review, the authors make only a good self comparison. They should be add a comparison with other methods (those cited in the introduction)

3) As a future work, I suggest to use the saliency map of the patchs as an input of the SSD method. This approach is already used to detect ground, aircraft, ship targets from SAR images. This issue should be added as future work and the following references can be added

a) Zhu, D., Wang, B., & Zhang, L. (2015). Airport target detection in remote sensing images: A new method based on two-way saliency. IEEE Geoscience and Remote Sensing Letters, 12(5), 1096-1100.

b) Tu, S., & Su, Y. (2016). Fast and accurate target detection based on multiscale saliency and active contour model for high-resolution SAR images. IEEE Transactions on Geoscience and Remote Sensing, 54(10), 5729-5744.

c) A. Karine, A. Toumi, A. Khenchaf, M. EL Hassouni, "Radar Target Recognition using Salient Keypoint Descriptors and Multitask Sparse Representation", MDPI Remote Sensing, May 2018, 10:843, May 2018.

Minor reviews :

1) A figure of the SSD architecture must be added in the Section 2

2) Avoid long sentences in the flowchart presented in Figure 2

3) Similarly of Figure 3, the source of Figure 3 mus be added in the caption. 

4) In figure 4, the axis titles sould be the same of those recorded in the link : http://worldweather.wmo.int/en/city.html?cityId=1501 . In the same figure, delete the double space after "and"

5) In page 7, replace "(CRESDA) (http://www.cresda. com/CN/index.shtml." by (CRESDA) (http://www.cresda.com/CN/index.shtml).

6) The authors say : "For all the Ori-results polygons (Figure 7(a)), we first obtained their center coordinate points, then conducted density-based clustering for these points (Figure 7(b))", it needs to add the name of density-based clustering method, DBSCAN, DENCLUE , .... ??? and add a reference

7) The acronym RPC has to meanings. This issue should be fixed.

8) In Section 4.1, the authors say "First, using the deep learning method, our method can automatically extract poppy parcel features without the need for manual selection and with a much faster detection speed." Can you explain this ? It is contradictory with what you say in ground turth part !

9) In Section 3.4, replace "W and H are the region length and width," by "W and H are the region width and length respectively,"

10) In figure 20, the "(c)" and "(d)" must be centered. Similarly for subtitile of Figure 21 and 22 and 14.

Author Response

Author Response File: Author Response.docx

Reviewer 2 Report

Dear Authors,

Please find in attachment the comments.

Good Work!

Comments for author File: Comments.pdf

Author Response

Author Response File: Author Response.docx

Round  2

Reviewer 1 Report

I suggest to accept thi paper.

It presents a good and chalenging work.

Reviewer 2 Report

Dear Authors,

In the present form, this article can be accepted.