Automatic Classification of the Ripeness Stage of Mango Fruit Using a Machine Learning Approach

Round 1

Reviewer 1 Report

Dear authors,

Please find below some comments and suggestions to improve your paper.

Kind Regards,

------

Title:

Add an “a”: Automatic Classification of the Ripeness Stage of Mango Fruit Using a Machine Learning Approach

 

Introduction

Lines 78-93: these lines describe methodology. Please remove or move to Materials and Methods.

 

General comment: more references on predictions of maturity in other crops using similar indicators need to be included. For example:

 

• Scalisi, A., & O'Connell, M. G. (2020). Application of Visible/NIR spectroscopy for the estimation of soluble solids, dry matter and flesh firmness in stone fruits. Journal of the Science of Food and Agriculture, 101(5), 2100-2107.

• Jha, S. N., Narsaiah, K., Jaiswal, P., Bhardwaj, R., Gupta, M., Kumar, R., & Sharma, R. (2014). Nondestructive prediction of maturity of mango using near infrared spectroscopy. Journal of Food Engineering, 124, 152-157.

• Shah, S. S. A., Zeb, A., Qureshi, W. S., Arslan, M., Malik, A. U., Alasmary, W., & Alanazi, E. (2020). Towards fruit maturity estimation using NIR spectroscopy. Infrared Physics & Technology, 111, 103479.

 

Materials and Methods

• Line 99: explain what DAFS stands for.

• Lines 111-115: the CIELab colour space has been proven to be better related to maturity in several fruti crops. Specifically a* and hue angle are two good indicators of fruit maturity in peach, nectarine and most fruit that turn from green to red. CIELab colour parameters can be derived from RGB. I suggest that the authors cite some papers where maturity was linked to colour even if in other crops:

  -Scalisi, A., Pelliccia, D. and O’Connell, M.G. (2020). Maturity Prediction in Yellow Peach (Prunus persica L.) Cultivars Using a Fluorescence Spectrometer. Sensors, 20(22), p.6555. https://doi.org/10.3390/s20226555

  - Scalisi, A., O’Connell, M. G., Pelliccia, D., Plozza, T., Frisina, C., Chandra, S., & Goodwin, I. (2021). Reliability of a Handheld Bluetooth Colourimeter and Its Application to Measuring the Effects of Time from Harvest, Row Orientation and Training System on Nectarine Skin Colour. Horticulturae, 7(8), 255.

  -Robertson, J. A., Meredith, F. I., Horvat, R. J. and Senter, S. D. (1990). Effect of cold storage and maturity on the physical and chemical characteristics and volatile constituents of peaches (cv. Cresthaven). Journal of Agricultural and Food Chemistry, 38(3), 620-624. https://doi.org/10.1021/jf00093a008

  -Ferrer, Ana, Sara Remón, Angel I. Negueruela, and Rosa Oria. "Changes during the ripening of the very late season Spanish peach cultivar Calanda: feasibility of using CIELAB coordinates as maturity indices." Scientia Horticulturae 105, no. 4 (2005): 435-446. https://doi.org/10.1016/j.scienta.2005.02.002

General question: Why have the models not been validated on external samples but rather just by four fold cross-validation?

General question: Why did the authors not use some portable NIR spectrometers (i.e., the most commonly used devices for assessments of mango quality and maturity attributes?) The methodology used in this study implies that in order to perform accurate predictions of ripeness, destructive measurements such as soluble solids and TA still need to be done after this study.

 

General question: Other commonly used methods for modelling fruit maturity are PCA / PLS regressions. Is there a reason why they have not been tested in this study?

 

 

Results

Lines 229-230: please rephrase to say that “the colour attributes used in this study (RGB) were not good indicators for classifying ripeness”.

 

Author Response

Please see the attachment.

Author Response File: Author Response.doc

Reviewer 2 Report

A well written paper.

Authors present ML approach to class fruit maturity of Mango.

 

Minor edits/suggestions:

Suggest replace 'after harvest' with 'at harvest',  as this is not a post-harvest (storage/handling) study.

Introduction (lines: 44-49): authors should consider benefits/limitations other published studies on mango fruit quality e.g. 'non-invasive assessments using near infra-red spectroscopy' and statistical approaches (e.g. PLS).

Line 99: define DAFS rather than in line 101.

Line 199: Figure 3: Authors need to explain the logic behind why the ripeness ML approach requires destructive sampling rather than state-of-the-art non-destructive method(s). Is there barriers to industry adoption for non-destructive method(s)? However, Fig 3 shows the Validation uses a non-destructive approach.

Line 226: Results: Insert a table of summary statistics of fruit data (ten variables: physical + biochemical + electrical properties) to provide readers with context of fruit quality and variability.

Line 232, Figure 4: Explain why colour change (RGB) detection can on be clearly detected using colour sensors?

Line 344: Table 8: A better title caption is required and describe/define each model (1. . 5). 

Lines 364 -384: delete if not required, otherwise add relevant statements.

 

Author Response

Please see the attachment.

Author Response File: Author Response.doc

Reviewer 3 Report

Overall I liked the work. Though the analytical steps used are typical for any Machine Learning based application, authors have applied them to the level of satisfaction. The only novelty of this manuscript is the domain area, i.e., predicting the "Ripeness Stage of Mango"; therefore, Figure 3 becomes all that more important. 

1. Is data primary? is it available for others to view ( or with limited views)?
2. Are biochemical physical and electrical characteristics ( pre-defined class of features) considered while using ML techniques? Are they thought the same and inputted to SMOTE?
3. Implementation details may be elaborated further.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Most of my recommended changes have been incorporated by the authors; this manuscript may proceed for further processing