Review Reports

Round 1

Reviewer 1 Report

The proposed article gives an insight into the possibility of using hyperspectral imaging for the detection of bitter almonds in sweet almond batches. Even though the topic is interesting there are some things that have to be addressed and explained in detail. It is said that the bitter almonds consist of  amygdalin. What are the usual ways of detecting amygdalin levels. This should be added in introduction. In order to judge about the novelty, one must firstly be completely informed about the problems that is being investigated.

Line 47-52: senteces is confusing and should be rewritten

Line 66: one bitter kernel is problem for the industry -since the article shows that at low concentrations of bitter almonds the proposed method is not effective I suggest rewriting this part

Sampling technique, imaging acquisition and processing, as well as classification models are well known from previous work.

Line 98: explain the origin of noise

In the Results section explain the differences between spectra of different almond groups  in more detail.

In Figure 1 correct the title on y axis and picture description.

Line 217: each Roi [Dana not shown]  -correct

LINe 236: explain what other features and why they are not useful

Line 240-245: add reference for peak analysis

Add y axis title on Figure 2.

Line 265-266: how were the data for amygdalin concentration obtained

The conclusion is too short, it should be expanded to give more specific conclusions. It is evident that data reduction does not give satisfactory results, and that different cultivars  show different challenges for multispectral analysis. This should be emphasized in the conclusion and in the abstract.

Author Response

Please see the attachment

Author Response File: Author Response.doc

Reviewer 2 Report

This manuscript detects bitter almonds in sweet almond batches. The overall quality is good. However, some issues should be addressed.

1. Why pixel-wise classification is used? It seems the intact almond can be used for classication. Thus, it makes no sense that using pixel-wise classification, especially, the ground truth samples are missing in the Figures. Ground truth image should be provided with marks.
2. What do you mean by Mixture 5%, 10%
3. It would be better to add a step convert the pixel-wise classification into object-wise classification for comparison.

Author Response

Please see the attachment

Author Response File: Author Response.doc

Round 2

Reviewer 1 Report

The manuscript has been altered according to the suggestions.

 

Author Response

All the suggestions were previously considered.

Reviewer 2 Report

Please carefully consider comment 1 and comment 3 in the last round of review. I prefer object-wise classification rather than pixel-wise classification, according to the objective of this study. 

Author Response

Please see the attachment

Author Response File: Author Response.doc

Round 3

Reviewer 2 Report

Thank you for the response of my comments. It is interesting that the authors indicate that ‘in the almond industry large flows of product are processed, so it is not possible to carry out an analysis of each individual kernel or to guarantee that the almonds are separated during the analysis.’ So how you label the samples to be detected and pick them out. A segmentation procedure is necessary. On the other hand, complex sorting systems should be developed under this situation. Also the object-wise study could also be used to explore the feasibility of the research objective of this study. The authors should conduct an in-depth discussion of this research, including the issues mentioned above.

Author Response

Please see the attachment

Author Response File: Author Response.doc