Image Segmentation of UAV Fruit Tree Canopy in a Natural Illumination Environment

Round 1

Reviewer 1 Report

The paper presents a methodological approach for unsupervised leaf classification in crops from UAV imagery. The document is well structured, the methodology and results section is accurately described, detailing the analyses carried out in the research. In general, some aspects of the introduction, materials and conclusion sections should be revised to make the justification and description of the research more complete. Below are some notes on the revised aspects of the research.

Line 37. One could also talk about pixel size and reading angle.

Line 48. El NDVI requiere de bandas que no están dentro del espectro visible (RGB) 

Line 67-68. These three options can be put in bullet points so that the existing options become clearer

Line 74-77. [39] It is not explained what the purpose of this segmentation is, doing so would help a better understanding of the work cited.

Line 77-80. [40]. What are the results? If the work is not adequately explained, it is not appropriate to cite it.

Line 80-82 [41]. Same comment as above

Line 97-99. The expression "In a word" does not seem scientific, I would look for other expressions such as “acording”, "in conclusion". This sentence should define what are controlled and uncontrolled conditions, it is difficult to know what is being referred to.

Line 112-113. The last sentence is not integrated in the paragraph, it should be introduced with something similar to "as a main conclusion" or "as a main result".

Line 114. This point could be materials and study area. Figure 1 should be contained in this item and referenced in the first paragraph.

Point Materials. This point is not very well defined, a photo of the field could also be introduced in which not only the spatial distribution of the crop but also the size of the trees, trunk size, crown size, leaf size, etc. could be shown. If a scale bar appeared in this photo, it would be ideal to get an idea of the dimensions. As for the UAV equipment, the size of the sensor has been defined but not the focal, resolution of the sensor is crucial to calculate the ground pixel size (GSD), as well as the GSD should be entered in the text.

The UAV itself does not produce orthophotos, does it work with the images from the drone, does it work with the orthophoto product of the survey obtained from a photogrammetric process, does it work with the orthorectified images after a photogrammetric process, does it work with the orthorectified images after a photogrammetric process? It would be interesting to answer these questions.

Line 127-138. To make this methodological definition, it is more convenient to use bullet points or graphics that show the work process.

Line 245-246. The first sentence is not explained, if the classifiers are not explained or belted, it does not add anything to the text.

Figure 5. figure b is mounted on figure a

Figure 6 two images are overlapping

Line 290-294. “….it is not difficult to find that the contrast…..”  The wording may mislead the reader

Figure 7 should be contained in point 4.1. 

Line 369-371. This sentence can be misleading, in this work the results obtained for two methods are used and compared, it would be necessary to define which method is meant in the sentence.

“The proposed method”, the wording of the document should be reviewed as it is not clear that there is an initially proposed method that will be purchased..

Line 402. The image size and pixel resolution should be indicated in the materials section.

Line 410. “(1) A shadow region luminance…” tis is not a conclusion “(2) The image segmentation experi-ments of the proposed…” is not a conclusion either.

Conclusions. the wording of this paragraph should be revised.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

The work has scope for improvement. Please improve with the suggestions as below:

1. Study area is not a point, it is an area -  --- change and represent properly in the figure and in the text.

2. Flight plan should be included in the methodology along with front and side overlaps or number of clicks/sec while UAV flying with 5 m/s speed.

3. why flying height is 20-60 m? why dont you fix the flying height?

4. How many days observations are taken? and their dates.

5. Atmospheric conditions of the flight days?

6. Oct2018- sunny/cloudy days is not sufficient ------ you must explain in section 4.4 in detail -- which days (date & time) images are consider and why ?

7. Explain adobe manual segmentation process also in one or two sentences.

Otherwise, work is good.

Author Response

Please see the attachment。

Author Response File: Author Response.pdf

Reviewer 3 Report

The objective of this paper is the development of an unsupervised image segmentation method, showing excellent quality of segmentation, high robustness and strong ability of generalization for extracting fast fruit canopy from aerial fruit tree images, imprinted by UAV in a natural illumination environment.

The article structure is satisfactory and the objective is clear. Each section (Introduction, Materials, The Fruit Tree Segmentation Method etc.) is analyzed in detail, while each aspect is explained properly. Moreover, a detailed mathematical, step-by-step background is provided in 3.2.1 and 3.2.3 paragraphs, clarifying the clustering results. The Results-Discussion section is coherent, while the numbered conclusions in the corresponding section highlight clearly the major points of the paper.  

However, the following corrections should be performed, which will improve the paper:

Paragraph 3.2.2: Although the clusters initialization is described in detail, only one bibliographic reference is included. Please, include the appropriate references, in order to set clear the mathematical background establishment.

Figure captions: All figure captions are modest and include only the necessary information. All captions should be enriched with more information, in order to be clear to the reader.

Comments for author File: Comments.pdf

Author Response

Please see the attachment

Author Response File: Author Response.pdf