Mapping Asbestos-Cement Corrugated Roofing Tiles with Imagery Cube via Machine Learning in Taiwan

Round 1

Reviewer 1 Report

The paper briefly describes a procedure for the detection of asbestos roof tiles in Taiwan.
The application is interesting, given the large and heterogeneous area investigated.

English language should be checked. Some hints are provided below, but they are not exhaustive.

In the paper the same method is called "deep learning classification" or "CNN" in different sentences and this can cause some confusion.


L25 "remove those buildings": which buildings? Buildings identified as having asbestos roofs in the aereal images but not in recent Sentinel-1 images?

L83-85 "Interventionary studies involving animals or humans, and other studies that require ethical approval, must list the authority that provided approval and the corresponding ethical approval code.": I am not sure why this statement is here and what is the link to the rest of the section.

L91 "Identify different roof types [...] have been made." -> "The identification of different roof types [...] has been achieved."

L150 "Then the portion images of building rooftops were reserved by joint the eCongnition feature exaction function and boundary detection method. ": I am not sure what "reserved" means here. Maybe "identified"? Or "partitioned"?

L153 "of not asbestos roof tiles" -> "of non-asbestos roof tiles".

L156 "Samples of applied imagery": used imagery? Available imagery?

L169 "automagical"? (I would delete this word).

L173 " (Hasan et al. 2019)" should be "[13]".

L195 "These types were defined as difficult pairs, as it was difficult to distinguish between them.": are them both covered by asbestos roof tiles?

L233 "both ANN and SVM method" -> "both CNN and SVM methods"

L234 "where is not detected by either" -> "where it is not detected by either".

L236 "were re-examined by either Google Street view or Unmanned Aviation Vehicle (UAV) images": by a human operator?

L240 "The asbestos roof exists in the historical aerial photo but no longer exists at present
day" -> "The asbestos roofs which exist in the historical aerial photo but no longer exist today"

L252 I do not understand this sentence: "Irregular boundaries polygon marked by the joint AI method then be expanded into the roof edge by the wavelet method for measuring area". Does it means that the wavelet method for boundary detection proposed in [18] has been applied to merge polygons boundaries detected by both methods to identify the roof edge and evaluate the roof area?

L269 How is "recall" defined?

L279-283 I am not sure of the procedure: asbestos roof tiles identified by SVM but not by CNN have been manually checked and corrected when necessary?

L319 "Wherever an asbestos cement corrugated roofing tile larger than four pixels (50  50 cm) could be detected.": either part of the sentence is missing or this is part of the preceding phrase, but then the full stop must be substituted by a comma.

L323 "missed judgments": false negatives?

L329 "and rule out the not existence building by compare the agreement from two temporal remote sensing data with machine learning" -> "and remove no longer existing buildings by comparing two remote sensing data sets with machine learning" (maybe no longer covered by asbestos roof tiles).

L336 "not to mention to acquire all the images within short period of time." -> "not to mention the difficulty (impossibility?) of acquiring all the images within short period of time."

L355 "Vertical imagery is the source of this study that could not detect the asbestos cement corrugated tiles lay aside the building, which is common phenomenon in the tropical region" -> "In this study nadiral or quasi-nadiral images have been used, therefore the detection of asbestos cement corrugated tiles on vertical walls, which are common in tropical regions, is not possible."

L359 "to detect its existence" -> "to detect their existence".

Author Response

 "Please see the attachment."

Author Response File: Author Response.docx

Reviewer 2 Report

The idea presented in this study is very interesting, and the efforts of contributing to reduce the use of asbestos fibers are much appreciated.

There are some corrections that must be introduced in the manuscript before I can recommend its publication.

In the attached commented PDF document, you can find, highlighted in green, some of the needed corrections that must be introduced at the English level.

The first paragraph of section 2 should be removed.

In lines 145-147 you state: “The ground-truth database had an accuracy of 90%, as validated through the independent inspection, and the sampled locations covered the entire island to guarantee the database completeness.” How is it possible to assume the accuracy you claim (e.g., at the conclusions section) if the ground-truth itself has a 90% accuracy?

In the conclusions section you should state the number of hours of computation time (line 349), and provide the link of the website (line 352).

Other minor corrections are also noted in the attached commented document; please refer to it.

Comments for author File: Comments.pdf

Author Response

 "Please see the attachment."

Author Response File: Author Response.docx

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.