Evaluation of a One-Dimensional Convolution Neural Network for Chlorophyll Content Estimation Using a Compact Spectrometer
, Rei Sonobe 2,3,*, Hiroto Yamashita 2, Haruyuki Seki 2, Harumi Mihara 1, Akio Morita 2 and Takashi Ikka 2,3Round 1
Reviewer 1 Report
The manuscript by Nofrizal et all is devoted to important applied task: comparison of efficiencies of using specialized spectrometer FieldSpec4 and simple spectrometer Colorcompass-LF for revealing concentration of chlorophylls on basis of analysis of reflectance spectra. The work seems to be interesting; however, there are questions and comments:
- “Keywords” should be improved; e.g., 1D-CNN, chlorophylls, reflectance spectra, etc. can be added.
- Section “2.1. Measurements and Datasets” should be improved. Now, it is not clear:
- What was age of investigated plants?
- Lines 148-149: Calibration procedure should be described in more details.
- What was total quantity of records?
- Figure 1 showed that detached leaves were used for investigation. This detaching can influence spectral properties of leaves (e.g., through induction of electrical signals or hydraulic signals). Why were detached leaves used? What was time interval between detaching and measurement? These points should be clarified.
- Section 2.4: What was criterion of significance used?
- Table 1:
- Significances of differences should be shown.
- Control variant had maximal concentration of chlorophylls. What were reasons of this result?
- Figures 3 and 4: Captions of figures should be extended; details of experiments should be included.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This article mainly introduces a system based on a cost-effective fingertip-sized spectrometer (Colorcompass-LF) . Aiming at the low cost vegetation properties estimation, the method mentioned in this article to estimate the chlorophyll contents of radish and wasabi leaves has a certain degree of innovation. And the effect is more robust and accurate than traditional methods. The quality of this paper is good. I recommend accepting this paper after minor edits.
- Abstract: should provide some quantitative results for the comparison and evaluation experiments.
- Model Development: the 1DCNN and DBN are existing models, I suggest the author to highlight their innovated technologies used in this study, and explain more on how it works better on the estimation of chlorophyll contents of radish and wasabi leaves.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Globally, the manuscript is very well written and organized. There are only some minor math style corrections that should be introduced in the final version of the manuscript, which are highlighted in the attached PDF file; please refer to it.
Additionally, and my only comment/suggestion, is the following one. The authors state in the abstract “(…) a system based on a cost-effective (…)” and in lines 52-53 “Consequently, the development of a low-cost hyperspectral remote sensing system would prove useful [39]”. I do agree with this last sentence, but the authors did not present a (provisional/foreseen) total price for the proposed solution (presenting also the expected/predicted price of all the individual components/ devices/ etc.). It would be very interesting to compare it with the prices of the existing solutions. As such, I recommend the authors to provide such comparison, perhaps by using one (or more) table.
Comments for author File: Comments.pdf
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
