Thermal Mapping from Point Clouds to 3D Building Model Facades

Round 1

Reviewer 1 Report (New Reviewer)

Comments for author File: Comments.pdf

The language of the manuscript needs minor corrections.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report (New Reviewer)

This paper proposes a framework to map temperature attributes from thermal point clouds to building facades to generate thermal textures for 3D analysis. Nearest neighboring point is used to project attributes from point clouds to texels. Overall the proposed method is simple and lacks novelty. Below are some detailed comments.

(1) The significance of the study should be further strengthen. What is the benefit of projecting thermal information from point clouds to 3D building model facades? Let me ask in another way. Why not directly using 3D thermal point clouds?

(2) Literature should be further organized. A lot of literature are listed in related work. However, it is not easy to understand the logic.

(3) Three different ways are introduced to calculate the intensity values and compared. However only one method is selected and employed. It is not necessary to list all the three ways in the methodology part, which may mislead readers. Instead the other two ways should be put in the discussion part.

(4) If a texture pixel in the 3D model is too large, the resolution will be very low. More importantly, a lot of thermal information is discarded from the original thermal point cloud. How do the authors explain this?

(5) How are thermal point clouds generated with MLS point clouds and thermal images?

(6) Manul registration is carried out between point clouds and 3D building models, which are time- and labor- consuming. In fact, the thermal information in a 3d environment is changing all the time. Automatic 3D thermal mapping is important.

(7) How does the registration error impact the mapping results. It should be discussed.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report (New Reviewer)

The authors have carefully read and elaborated on all the instructions. They have evaluated all the recommendations and implemented or refuted them as necessary. Please see the attached file. 

Comments for author File: Comments.docx

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

The paper presents a method of mapping TIR point clouds to 3D building model facades to establish an extended 3D thermal description.

In the paper, it was mentioned that the point cloud and the building model are already co-registered. Therefore, the work presented in this paper only contains texture mapping that does not involve any coordinate transformation. The innovation presented in this work is considered limited. Additionally, the experimental parts are insufficient because the necessary efficiency and accuracy analyses were left for future research.

Reviewer 2 Report

General Description:

The overall subject of the reviewed paper is interesting, but there is a clear lack of contribution to the state of the art. The paper lacks an appropriate quality of academic writing. The paper contains punctuation errors, missing articles, and completely incomprehensible sentences. The paper is not or only slightly following the general content structure of an academic publication and should be revised carefully (e.g., abbreviations are not explained first before use). In “results and discussion”, the authors pick up their own previous case study without making this very clear and use this data to investigate the three proposed methods. The methodology, discussion of the results is incomprehensible, and it is not communicated precisely which methods yield the best results and for which reason. The outcome of the paper seems to be that windows are cut out from a texture map after which a ghost filtering is applied on thermal images. Overall, we advise the authors to continue working on their method until a sufficient contribution can be achieved.

Specific remarks per section:

Abstract:

The Abstract appears lengthy and tedious before getting to the point. The preferable structure should follow the general – problem statement, proposed solution, results - logic, before going into detail about the problem of the Sota. More concrete information is necessary about the “three different ways to calculate the intensity values”. Which solutions are proposed, how was their performance assessed and which one was the best performing? What is this meaning in the field i.e. the contribution?

1.0 Introduction:

Starts from a different subject. Climate change is far off topic. The problem statement is not in connection to the proposed method of the paper and very general. Additionally, clear contributions are lacking.  

Related Work

 

there is no clear overview of what methods have been proposed before, what the state of the art is and what challenges are faces today. In general, it is not clear to the reader what the main issue is and why it is not solved yet by previous research.

Preliminary Work

This subjection finally explains the related work in the field and demonstrates the drawbacks of the existing sota. However, the precise subject is about the lack of detail in thermal textures. Unfortunately, it is not made clear why this is the case. Several times it is mentioned that the "best textures" are selected, which does not describe the methods of the mentioned studies or the hurdle.

 

L132: This sounds positive and does not seem to make sense, please explain better how lower resolution is improving discretization. What does it mean if the number of matches decrease?

Recommendation: Combine 1.1. and 1.2 as “2 Related Work” and give a clear overview why traditional methods are not ideal, what is a general research outline (how to get a 3D model, what is possible) and what is the current issue. What has been tried to get better textures. Have other researchers tried the 3 methods you are proposing? Discuss their and other methods results?  

1.        Methodology

It is not clear which part of the method is novel about this approach and therefore it seems rather trivial. Probably, this assumed misunderstanding can be improved by a better related work section. Finally, some potentials and drawbacks of different methods are discussed. However, this in fact is not the right position in the paper, as it will be applicable to previous research as well. Moreover, the illustration of the section lacks professionality (i.e., way too large or unreadable small text, different image sizes etc.). The methodology is strongly based on previous work of the authors (Zhu, J.; Xu, Y.; Ye, Z.; Hoegner, L.; Stilla, U.: 2021) and it is not made clear in which way this paper has any contribution towards improving the existing method.

 

2.        Results in discussion

 

The section starts with explaining the test case. However, it remains incomprehensible how big the test area is. Additionally, it is clearly just a repetition of the author’s previous work (“Thermal point clouds were provided by [5], who used laser scanner point clouds and TIR image sequences of facades captured from a mobile platform setup [33]”). Despite clearly stating otherwise in the methods section, they start with LOD2 models in the described case study. Therefore, subsections 3.1 to 3.3 explain how to get to LOD3 models, but this can hardly be called “results and discussion” and should be moved to the methodology section. In 3.5, finally the results of the 3 thermal mapping approaches are shown. However, the choice of the evaluation criteria is not clear. The discussion of the results is very poor, and it remains incomprehensible which method performs best.