Next Article in Journal
Differential Privacy-Based Spatial-Temporal Trajectory Clustering Scheme for LBSNs
Next Article in Special Issue
A Road Crack Segmentation Method Based on Transformer and Multi-Scale Feature Fusion
Previous Article in Journal
Power Flow Management of Interconnected AC Microgrids Using Back-to-Back Converters
Previous Article in Special Issue
Study on Parking Space Recognition Based on Improved Image Equalization and YOLOv5
 
 
Article
Peer-Review Record

Multispectral Remote Sensing Image Change Detection Based on Twin Neural Networks

Electronics 2023, 12(18), 3766; https://doi.org/10.3390/electronics12183766
by Wenhao Mo 1,*, Yuanpeng Tan 1, Yu Zhou 2, Yanli Zhi 2, Yuchang Cai 1 and Wanjie Ma 2,*
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Reviewer 3:
Electronics 2023, 12(18), 3766; https://doi.org/10.3390/electronics12183766
Submission received: 17 July 2023 / Revised: 25 August 2023 / Accepted: 30 August 2023 / Published: 6 September 2023
(This article belongs to the Special Issue Computer Vision for Modern Vehicles)

Round 1

Reviewer 1 Report

This manuscript proposes a change detection method for multispectral remote sensing images using a Siamese neural network involving feature extraction from dual-temporal remote sensing images based on a ResNet-18 network. The authors designed an attention module network structure to enhance the feature extraction and utilized an adaptive threshold comparison loss function to solve false alarms in change detection for more data sensitivity and robustness. The enhanced algorithm framework was evaluated via experiments, feature visualization outputs, and false alarm situations, demonstrating the overall performance enhancement.

 The paper is scholarly written and organized, highlighting its novelty and contribution, performed in-depth literature on the subjected area and adequately referenced, clearly discussed the approach, and performed comprehensive evaluation, analysis, and comparison of the proposed algorithm framework and presenting the performance improvement through various experiments and visualizations.

 Based on this review, I, therefore, accept the paper in its present form.

Author Response

Thank you for receiving my paper

Reviewer 2 Report

Overall the paper is decent, however the results are not very strong nor desired focus of this change detection is outlined.  A few comments:

1) What is the goal of the change detection of the land objects? 
2) For the small amount of data, why is data augmentation not utilized? 
3) The results of the training is not well presented? 
4) Is the developed model available for the general community to utilize? 
5) Can we add a confusion matrix to analyze these results? 
6) What are the next steps and future work for this model? 

Written in clear English, no major issues found

Author Response

Thank you for receiving my paper

Author Response File: Author Response.pdf

Reviewer 3 Report

This manuscript focuses on the study of multispectral remote sensing image change detection based on twin-neural networks. This manuscript needs to be well revised and improved before publication. Some suggestions to improve the paper:

(1) Line 30-32 and Fig. 1: I suggest moving Fig. 1 to the Section 4 “Dataset”, and the used wavelength bands (~nm?) should be presented with the figures. Besides, “Figure 1shows” -> “Figure 1 shows”.

(2) Line 59: “In order to” -> “in order to”.

(3) Line 70: The full name of STAnet should be added.

(4) Line 91: The full name of FCN should be added.

(5) Line 115-116: “high resolution” -> “high spatial resolution”.

(6) Line 123:  The full name of OTSU should be removed from Line 234 to here.

(7) Line 130: The full name of SAR should be added.

(8) Line 149: “LI” -> “Li”.

(9) Line 158: The full name of ResNet-18 should be added.

(10) Line 187: “Pyramid” -> “pyramid”

(11) Figure 2, 3, 4: The representations of 3 figures are unclear and ambiguous, which needs to be further polished, and the symbols used in these figures should be well explained. Besides, I suggest adding a general flowchart to illustrate the methodology and methods used in this paper, which is very important.

(12) Line 206: There in not the so-called No. “Formula 3.1”.

(13) L211-213: Since these 3 sentences are an explanation of formula (1), which should belong to the last paragraph.

(14) Section 4: The presentation for “DataSets” is incomplete, some details should be explained: 1) What kind of image values are used in this study for change detection, DN, radiance, TOA reflectance or surface reflectance? 2) How many multispectral wavelength bands are used? in which, the spectral range should be listed. 3) Since the change detection based on precise pixel-to-pixel matching, has the dataset cover the geometric rectification?

(15) Line 300 and Fig. 6: The meaning of “Nir” and “rgb” should be explained.

(16) Line 348: “5. Conclusions” -> “6. Conclusions”

(17) Line 349-350: “In order to …mor quickly”, how quickly is quickly? Therefore, some discussions and comparisons should be added to show the “quickly” of your method developed in the paper, which also is indispensable for this study.

Minor editing of English language required

Author Response

Hello editor.

I have highlighted the modified areas in yellow. Please review them. Thank you.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Please double check Line 246-253, why list the (1), (2), (3) with blank (1) and (3)?

 Minor editing of English language required

Author Response

I have revised,tank you.

Back to TopTop