Maize Stem Contour Extraction and Diameter Measurement Based on Adaptive Threshold Segmentation in Field Conditions

Round 1

Reviewer 1 Report

In the paper authors describe an imaging  method to measure stem, (particularly the second internode), diameter on maize plants grown under open field conditions. Authors used a handy RGB-D camera to take images in the field, and images were elaborated by a pipeline that combines HSV + Otsu algorithm. Stem diameter was finally estimated by a "reference method" that compares stems width with a standard checkerboard of known size used in images. The method was intended, according to the authors, to overcome some shortcomings of RGB imaging, such as light conditions, or manual measurements. 

The paper present many weakness in terms of structure, description and technical approach that must be addressed.

The title is somehow misleading. The term "In situ extraction of Maize Stem Contour...." must refer to the fact that the overall pipeline is carried out directly in the field". However, reading the paper there is no evidence or statement that confirms this element. It seems that the in situ should refer to the imaging acquisition phase only, since the rest of the pipeline was likely implemented in the office. It is recommended to use "open field conditions" instead of in situ, more appropriate for the study.

The beginning of the introduction (line 30-line 41) was elaborated with very short statements pointing at specific references, with no clear glue that puts them into a logic reasoning. If the point was to affirm the importance of plant phenomics in breeding, it should be elaborated in a better and more fluid way.

In the introduction (line 45-line 82) authors introduce elements of comparison with other imaging methods that were recently used to measure plant stem our trunk diameters, stressing the importance to carry out these type of approaches directly under real open field conditions. Whilst this hold true for precision farming (for example estimating yield) and plant breeding (i.e estimating lodging resistance), it must be pointed out that plant phenotyping by imaging is also relevant in plant science where controlled conditions are very important.

In the material and method section, I suggest to eliminate the sentence on line 93-95, a repetition of what was already stated in the previous paragraph from line 87 to line 91.

The material and method section poorly describes the experimental setup regarding the image acquisitions. More information on the experimental field (plant density, plot size) must be provided, elements relevant for canopy shadows. The geometry of the imaging set-up is missing  (distance of camera from plant, height of camera, use of tripod or manual shooting (?)) Only camera angle (from vertical or horizontal?) is indicated (line 101). An image of camera set-up can help in understanding.

No indications are present on how the imaging areas and plants were chosen (random?) and shooting time and hours have not been indicated (important for sun position and light quality). 

Imaging methods are properly described.

The reference method described starting at line 173, refers to a checkerboard grid consisting of black and white squares of known size used as a reference object. Not clear if it has been used for any of the acquired image as a reference. Positioning of the reference grid respect to the observed object, is tricky. Since the error is in the size of mm, the alignment of the grid with the plant is crucial (vertical and horizontal alignment). Looking at Figure 5a, the grid look inclined respect to the vertical line and the vertical behavior of the plant.  

The method of stem measurement must be better described. It is not clear how the region of interest (second internode) is identified (manually?, how the pint of the contour are selected in the extracted image, how many measure are carried out (it seems from Figure 5b that three different diameter are calculated per each image). Overall it must be clear if the pipeline is automatic or requires some manual steps.

In the result sections, the data are presented in a clear way and the comparison of the three different pipeline used is well documented. 

The discussion in part is not consistent with the results. Authors comment some shortcomings (from line 242) due to light quality and conditions, that were not introduced by results. Results should  present data on the variability of the method due to light condition. According to their statement, it means they presente data form the "best" pictures", if so they should comment this point. It is also clear that this shortcomings are the same that authors criticized in the introduction for other imaging methods to partly justify the development of their approach.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The manuscript compares different imaging processing techniques for contour extraction of maize stem. The manuscript is very well written but the English in the abstract needs to be re-viewed.

The abstract should also briefly state the significance of this work.

Line 20 of abstract need to indicate the algorithm of choice given the study results.

 Line 219:  mention the proposed algorithm

Line 223: mention the proposed algorithm

Line 235: mention the method

Line 241: mention the method

The conclusion must name the proposed superior algorithm and not just implied.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Authors properly addressed all the comments. The answers they provided and the integration are satisfactory. The paper has been restructured according to the criticism and sufficently integrated in the introduction and method sections, providing a clear explanation of the methodology and approach implemented.

Author Response

Dear editors and reviewers：

We appreciate for your warm work earnestly, your suggestion makes our manuscript more specific and scientific, and greatly improves the quality of our manuscript. In addition, we have added the manuscript as suggested by the editor. The manuscript has exceeded 4,000 words. All additions to the manuscript are marked up using the "Track Changes" function. For your clarity, supplements to the paper were marked in green.

Kind regards,

Jing Zhou

Author Response File: Author Response.pdf