A Complete Environmental Intelligence System for LiDAR-Based Vegetation Management in Power-Line Corridors
Round 1
Reviewer 1 Report
91 DFIG -Data Fusion Information Group
285 Here, the average absolute accuracy equal to 19cm was measured, with maximal error equal to 37cm – accuracy of what? Error of what?
Chapter 3.2 – Statistical (percentual) evaluation of results and system reliability mises me, or/and numerical summary in compact table.
Author Response
We thank the Reviewer for their valuable comments and contributions to the quality of the paper. We have made considerable efforts to address all the reported concerns. The following includes the point-by-point response to the reviewer’s comment.
Response to Reviewer 1 Comments:
91 DFIG -Data Fusion Information Group
Response 1.1: Thank you. Corrected!
285 Here, the average absolute accuracy equal to 19cm was measured, with maximal error equal to 37cm – accuracy of what? Error of what?
Response 1.2: Accepted. We now state explicitly that we address encroaching vegetation for clarity.
Chapter 3.2 – Statistical (percentual) evaluation of results and system reliability mises me, or/and numerical summary in compact table.
Response 1.3: Accepted. As also suggested by other Reviewers, the results were extended by performing tests on different computer systems, where different processing times were measured, while identical results were obtained in terms of the proposed methods’ accuracies. Although the latter are, therefore, not reported separately, they prove the system’s reliability. Accordingly, this has been noted in the Discussion.
Author Response File: Author Response.docx
Reviewer 2 Report
This paper aims to build a processing framework for monitoring, predicting the situation of vegetation in powerline corridors and then providing the assessment of threat about it. In all, the contribution of paper is piratical. Nevertheless, there still exists some problem as view as my point.
1). The framework introduced seven levels of the whole process, however, only the part of level 3 introduced the details of processing, which should have been expressed by a series of math equations. Specifically, the level 0: data processing should include detailed method in its math expression, such as the method of generation of DTM from LiDAR. A confirmed method should be pointed out straightly.
2). No correlation can be extracted between the parameters introduced by Table2 and the prediction and monitoring of vegetation. If any model including these parameters was employed, please point it out straightly in the part of method and materials.
3). This paper only chose one studied area, if the framework is built for solving this sort of problem, more areas should be included for making it solid.
4). If the author aims to confirm the execution times of different algorithms (Table.3), more types of processors was suggested to be further tested.
In all this paper is practical and relatively well organized. However, the main problem is also obviously, the methods and processes should be expressed detailly!
Major revision is suggested.
Author Response
We thank the Reviewer for his valuable comments and contributions to the quality of the paper. We have made considerable efforts to address all the reported concerns. The following includes the point-by-point response to the reviewer’s comment.
Point 2.0: This paper aims to build a processing framework for monitoring, predicting the situation of vegetation in powerline corridors and then providing the assessment of threat about it. In all, the contribution of paper is piratical. Nevertheless, there still exists some problem as view as my point.
Response 2.0: We thank the Reviewer for his comments. As reported in the continuation, we have addressed all the issues exposed.
Point 2.1: The framework introduced seven levels of the whole process, however, only the part of level 3 introduced the details of processing, which should have been expressed by a series of math equations. Specifically, the level 0: data processing should include detailed method in its math expression, such as the method of generation of DTM from LiDAR. A confirmed method should be pointed out straightly.
Response 2.1: Agreed. As level 0 addresses data pre-processing, it was moved to the study area and data preparation section also (in accordance with the comments of Reviewer 3), however, the mathematical framework has been extended throughout all other methodological subsections, as suggested.
Point 2.2: No correlation can be extracted between the parameters introduced by Table2 and the prediction and monitoring of vegetation. If any model including these parameters was employed, please point it out straightly in the part of method and materials.
Response 2.2: Accepted. Although we noted in the last sentence of the first paragraph of Section 2.2.4 that the features defined by Table 2 were used as contextual features, their use was not explained clearly. Accordingly, we now state explicitly the purpose of these parameters, and provide a mathematical definition of their usage. In addition, Figure 5 has also been updated for clarity.
Point 2.3: This paper only chose one studied area, if the framework is built for solving this sort of problem, more areas should be included for making it solid.
Response 2.3: We understand the Reviewer’s concern, however, due to the Confidentiality Agreement with the ground-truth data provider, that is the national power transmission system operator ELES d.o.o. which operates 100% of power transmission lines network in Slovenia, we are unable to share additional results. In particular, as the national power transmission line network is considered to be a critical national infrastructure by the law, we have restricted access to the past activities related to the corridor management tasks (namely, dates, times, extents and detailed types of vegetation clearance tasks), which should be considered during the learning, as well as the accuracy assessment. Still, we now state explicitly at the beginning of the last paragraph of the paper that “the reported study provides only the experimental validation, while additional test areas need to be included during the system operation in order to achieve its demonstration in an operational environment” for clarity.
Point 2.4: If the author aims to confirm the execution times of different algorithms (Table.3), more types of processors was suggested to be further tested.
Response 2.4: Accepted. Tables 3, 4 and 5 have been extended and commented on accordingly.
Point 2.5: In all this paper is practical and relatively well organized. However, the main problem is also obviously, the methods and processes should be expressed detailly!
Major revision is suggested.
Response 2.5: We thank the Reviewer for his comments, which contributed greatly to the quality of the paper.
Reviewer 3 Report
1) Different stages of the process are highlighted. But, results to validate the necessity of those steps must be included.
2) From the second paragraph of the introduction section, the authors must perform a critical review on earlier literature.
3) What is the need for fusion? What information do you feel are missing in both the modalities of data which will be fused?
4) Many engineering methods are dealt in various levels. But, mathematical equations are missing. It must be included.
5) If any website is available for the database, it can be included.
6) What is the nature of input data used for level 0? some results must be included. Do you really think that this pre-processing is necessary? How do you validate it?
7) what are the different objects you are trying to assess in "object assessment" step? How will the results vary for objects of different sizes and shapes?
8) Does the power line scenario used in level 2 applicable for all geographic locations?
9) What is the necessity for k-NN classifier in regression model? Do you categorize the pixels with it? If yes, why haven't opted for state of the art methods?
10) Results section seems to be relatively weak. Individual stage results must be included
Author Response
We thank the Reviewer for his valuable comments and contributions to the quality of the paper. We have made considerable efforts to address all the reported concerns. The following includes the point-by-point response to the reviewer’s comment.
Point 3.1: Different stages of the process are highlighted. But, results to validate the necessity of those steps must be included.
Response 3.1: Accepted. In accordance also with the comments of other reviewers, the tests were extended by including different computer systems, while additionally considering the evaluation of the first two levels of data fusion. Latter are now reported in Section 3. 3.
Point 3.2: From the second paragraph of the introduction section, the authors must perform a critical review on earlier literature.
Response 3.2: Accepted. The introduction section has been extended accordingly.
Point 3.3: What is the need for fusion? What information do you feel are missing in both the modalities of data which will be fused?
Response 3.3: Thank you for the suggestion. We extended the first paragraph of Section 2.2 accordingly.
Point 3.4: Many engineering methods are dealt in various levels. But, mathematical equations are missing. It must be included.
Response 3.4: Accepted. Mathematical framework has been extended accordingly.
Point 3.5: If any website is available for the database, it can be included.
Response 3.5: Unfortunately, the data used in this study is under restrictive confidentiality agreement with the end-users, that is national power transmission system operator ELES d.o.o. that operates 100% of the 400 kV, 220 kV, and 110 kV transmission lines in Slovenia. As these are considered to be a critical national infrastructure by the law, releasing detailed information about their operations in a form of an open data would to be a risk to the security (as explained in the answer 3 to the comments of the reviewer 2). We are, therefore, unable to publish it.
Point 3.6: What is the nature of input data used for level 0? some results must be included. Do you really think that this pre-processing is necessary? How do you validate it?
Response 3.6: We agree with the reviewer that pre-processing step does, in-fact, not necessarily fit into the methodology section. Still, we believe it is needed for explaining how ground truth data was obtained. Accordingly, we moved this part of the text to Section 2.1 that addresses “Study area and data source preprocessing”.
Point 3.7: what are the different objects you are trying to assess in "object assessment" step? How will the results vary for objects of different sizes and shapes?
Response 3.7: Accepted. We now clarified at a beginning of the section 2.2.1 that object assessment deals with extraction of individual trees and fusion of auxiliary data sources about the environmental conditions, relevant for prediction of their development. Accordingly, their delineation is achieved with a sufficient accuracy, using a single tree-crown approach that we proposed a while ago [31]. Thus, we additionally note in the discussion that assessing the accuracy of tree-crown delineation on the growth prediction accuracy requires further study and will be considered in our future work, in the context of improving prediction accuracy.
Point 3.8: Does the power line scenario used in level 2 applicable for all geographic locations?
Response 3.8: Yes, providing the thresholds for filter definition are adjusted according to the legislation and other relevant environmental factors at a given geographic location. We thank the reviewer for the comment, this is now explicitly stated in the paper.
Point 3.9: What is the necessity for k-NN classifier in regression model? Do you categorize the pixels with it? If yes, why haven't opted for state of the art methods?
Response 3.9: Accepted. As we do not use k-NN classifier, it’s mentioning was indeed confusing. For clarity, we removed the referred statement.
Point 3.10: Results section seems to be relatively weak. Individual stage results must be included
Response 3.10: As noted in Answer 3.1, additional tests were performed on different computer systems, while all sages (levels) of data fusion are now tested separately. For clarity, the relations between the tests and data fusion levels (stages) are now additionally explained at the beginning of Section 3.
Round 2
Reviewer 2 Report
All problems has been revised.
Reviewer 3 Report
It can be accepted now