Optimization Method of Airborne LiDAR Individual Tree Segmentation Based on Gaussian Mixture Model
Round 1
Reviewer 1 Report
1. Lack of references, such as the first paragraph in INTRODUCTION.
2. Literature review is insufficient. Do not just simplely list some studies. Why authors chose those studies? What are general reseach objectives of previous studies? What are their contributions?
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
The paper entitled “Optimization Method of Airborne LiDAR Individual Tree Segmentation Based on Gaussian Mixture Model”, presents an individual tree segmentation optimization, based on Gaussian mixture model for airborne LiDAR data.
Results look quite impressive in terms of recall and accuracy and also the improvement compared to other methods are of interest.
The manuscript is well written and consider all formal aspect. The main concern is about the method. It is complex and composed of several steps, anyway it is quite difficult to follow. I suggest to rephrase it or some part of it for an easier understanding.
Line 10-11. Consider to rephrase, adding potential of Lidar to detect forest, that is useful for the following aspect.
Line 31. consider to remove “Known as the lungs of the earth”. The proper function of lungs is use oxygen instead of produce it.
Study area, consider to show where the dataset are come from.
Figure 1. Consider to normalize the point cloud and change “Coord. Z” with height and add morphology information in the main text
lines 133-134. Is it flat and at see level or it is normalized?
Line 146, please be consistent with LiDAR
Lines 152-155. What is the reason to use only 10 plots in the analysis?
Please, add flight information about Trento survey
Although complex, the methods appear really difficult to understand. Please, try to make it easier for the readers. Consider to add subparagraphs in conjunction with flowchart (Figure 2).
Lines 436-442 It is a method.
Figure 11. a-c, add height measure unit (better in m).
15 Figures seems to much.
Consider to move some of them to annex material
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
This paper proposes an ITS optimization method for airborne Lidar data based on the Gaussian mixture model. MS is used to pre-segment the tree point cloud directly, SVM is used to identify the abnormal segmentation samples, and SOM is used to classify the wrong samples. Finally, the error sample is segmented by the Gaussian mixture model. The article's structure and presentation are clear, but some technical details still need to be reconfirmed. Some scientific questions deserve further consideration by the authors:
1) The selection condition and range of bandwidth hr and hs in formula (1) are not given. I think bandwidth is a critical problem that needs to be elaborated.
2) What is the meaning of Crown Cap and Crown Cover in lines 8 and 9 of Table 2? The crown cap means the upper-most part of the tree crown.
3) The instability of formula 17 to optimize tree crown segmentation need to be elaborated in the Discussion section.
4) The proportion of tree species (coniferous forest, broadleaf forest) in the two data sets can be listed in Table 1.
5) The color bar of the tree height in Fig. 1 could be changed as the absolute heights of the trees in the experimental sites.
6) What does the number of cells in FIG. 5 and FIG. 14 represent? Is it the number of neurons?
7) What is the meaning of the color bar in Figure 7?
8) Please explain it in the caption of the figure with aligned and plausible magnitude ranges.
9) A similar work employs mean-shift and deep learning networks for tree crown segmentation, named “Individual Tree Crown Segmentation and Crown Width Extraction From a Heightmap Derived From Aerial Laser Scanning Data Using a Deep Learning Framework”, which should be mentioned in the Introduction section.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Dear Authors,
Thank you for the carful revision of the manuscript. I still have some little concern, mainly related to graphical aspect.
Be aware about citation. E.g. line 55 Aaron et al. [x], instead of point the number of the reference at the end of the sentence.
Consider to change pts/m2 with pts m-2
Figure 1 looks better. However, I intended the point clouds normalization, not the subtraction of min value to all the stand. Please check the values. There are not trees so tall in Italy.
Figure 1 caption, consider to remover the “experimental area” repetitions.
Table 1. Add space in Crown area (m…
Line 147. The added sentence is not needed. It concept of forest vertical structure is redundant, consider to remove it.
As for figure 1, please normalize the point cloud. Values presented are meaningless
Reference 50, the name and surname have to be changed. Paric C., Valduga D., Bruzzone L.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
I feel that the revisions and additions to the manuscript add value to both its readability and scientific merit. The authors did a nice job of tree crown delineation based on the airborne laser scanning technique. Therefore, the decision of acceptance was made.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf