AdTree: Accurate, Detailed, and Automatic Modelling of Laser-Scanned Trees
Round 1
Reviewer 1 Report
Contribution even in the submitted form is innovative, bringing new knowledge for the segmentation of cloud points and creation of tree models. It makes use of, integrates and complements existing knowledge from current and relevant known solutions.
The goal of proposing an approach to the correct and accurate reconstruction of a tree branch from a cloud point is timely, ambitious and challenging. Although the proposed method is based on relevant capping procedures, it is well documented and explained, the conclusions that it is functional and effective are not fully justified, and I do not consider the results sufficiently and perfectly interpreted.
There are considerable reserves in presentation quality. From a technical point of view, the contribution is written in an understandable style, as well as its structural breakdown in general. Quality and understanding of the principles supporting the included visualizations. The proposed segmentation method and partly the results are also presented appropriately. The work contains most of the essential parts required in the scientific publication - a balanced analysis of the issue and identification of unsolved, respectively otherwise solved problems, clear definition of the goal. The reserves also include an assessment of the quality of the results achieved and their comparison with the results of other authors in the field. In the narrower sense, some passages require transfers (section 3.1. is part of the methodology), the validity of the evaluation statements in the results and conclusions section needs to be supported by appropriate citations. The shortcomings mainly relate to the methodologies used and the results achieved.
From the formal point of view, the definition of the methodology of the work as such is completely absent and the requirements for the structure and content of the work of a developmental, programmer character are not fulfilled either. Interactive approaches or agile methodologies are appropriate for the development of smaller and specialized applications. However, the analysis of requirements and application specifications always play a key role. Requirements should cover technical, functional and user needs. It can be accepted that most of the relevant information in this regard is cited by the authors in the introduction, and a description of the proposed solution, but the formal requirements are not fulfilled. It is also necessary to define the architecture, system design, module structure and input and output data requirements in the specification. The results must then be demonstrated to meet the specified (technical) requirements. For testing purposes, I find it appropriate to design an appropriate experiment.
Although the authors define the requirement for the proposed segmentation tool to create geometrically and topologically accurate and correct models, they do not describe the proposed solution itself, and in particular its testing method. In particular, the issue of test data sets, together with the methodology and results of verifying the proposed procedure on them. Commenting on visualization of individual examples of solving partial problems (also creating complete tree models) is useful for understanding the principle, but it cannot replace the commentary of the functioning experiment results. Perhaps it would be sufficient to focus on the marginal positions of the diversity of tree types mentioned in line 248, (which is much wider than used by authors) - deciduous, coniferous, typical tree species (from point of view of branching), small (young), big (old) or to test data obtained from real conditions (eg Remote Sens. 2019, 11, 211; doi: 10.3390 / rs11020211).
While the correctness of the procedures used is indisputable, but with the above-mentioned objections to the way in which the proposed solution is valid. The study is technically sound, uses well-established and generally accepted approaches. It is also reproducible. However, it does not allow to formulate and comment on scientific results.
After completing and proving the validity of the conclusions, the results may be very interesting.
The text is comprehensible and written in a suitable style. But I do not feel competent to assess the level of language.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This paper addresses the problem of modelling tree trunks and branches from point clouds. The same problem has been object of interest before, and some solutions have been provided. In fact, there is some dedicated software, for instance PypeTree (Delagrange et al., 2014), that is not cited in this paper (although the authors cite a previous work of Verroust and Lazarus, 2000). TreeQSM (Raumonen, 2013) and SimpleTree (Hackenberg et al., 2015) are other free programs for single tree modelling not mentioned in this paper. These programs should be cited and compared, in some way, with the approach proposed by the authors.
The contribution of this paper seems to be in Section 2.2.2: the method to remove adjacent vertices and create new edges, that basically reproduces the Douglas Pecker algorithm. The method is outlined in Figure 9, but I can´t follow it. Is the new vertex between v1 and v2? What is a subtree? How do you define it? How do you model the branches that leave from node 2? Don´t you need to reconstruct the structure of the tree after this simplification. Table 1 shows variables to measure geometrical accuracy. The measure global accuracy, but what about local accuracy? What happens where the branches leave the trunk? How manage the algorithm the small branches? Which is the limit?
All these questions should be answered before the paper could be published.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
I have checked revised version of the paper Shenglan Du , Roderik Lindenbergh , Hugo Ledoux , Jantien Stoter , Liangliang Nan following also comments in Author response to my comments/questions.
I am fully satisfied with author s explanations and also corrections/improvements performed in original version. The manuscript has been significantly improved and now warrants publication in Remote Sensing.
