A Method for Turning a Single Low-Cost Cube into a Reference Target for Point Cloud Registration

Round 1

Reviewer 1 Report

A die-based automatic point cloud registration method is proposed in this paper.

The method is specialized only for close-range (up to 12m) laser scanning.

Claims made in the conclusion support the experiments and results.

A dice is required to be seen by all the scanners all the time?

Please explain if the system registers point clouds during scanning or if this procedure is performed at a later stage.

Master and slave point clouds acquisition is not explained in the text. I think the related details should be added in the experiment section.

Line 281-282: what do you mean by fitting two point clouds before/without registration captured at two different stations?

 

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Interesting article. Well written. The problem is clearly stated. Experiments and results could be improved a bit. Also, the contributions can be emphasized more. 

Considering just one target, the registration accuracy is function of the its size. It is expected that a multi-dice version would perform better in this regard with a good distribution in the scene. 

- What is the cost of the targets compared to a conventional scanner, like the Trimble SX 10 used?

- The proposed method has more computational complexity than with spheres? 

- Besides the RMSE and smaller cost of the targets, what are the practical benefits of the proposed method? How faster it was to set up and scan the scene? 

- The distance between scanner to target gives some insight. However, it would be better if we had some average point density of the scanned target, and how this metric affects the RMSE. 

 

Minor comments:

line 104-107 - A solid with more facets would eliminate this requirement ... dodecahedron for instance.

line 305 - Doesn't need to be an abrupt temperature change. A significant change in temperature during the scan?

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Dear Authors:

    Your article: “A Method for Turning A Single Low-Cost Cube into A Reference Target for Point Cloud Registration” presents a very relevant theme for studies that use a cloud of points obtained by static laser scanner systems. Even this condition can be better explored in chapter 1 as the alternative of backpack laser scanner systems but does not have comparable accuracy. The wording of the article requires proofreading by a native English speaker and the thirty-two bibliographical references should be explored in the introduction and discussion of the results. Another aspect that must be justified is the performance of only two scans for each dataset. I have presented sixty-two comments of minor importance that can be verified in the digital file (comments) and I would like your special attention to the following observations:

Abstract: Present the place where the study was developed.

Chapter 1: highlight the objectives of the study (general and specific)

Line 137: Introduce the concept of chirality.

Present the hardware resources (personal computer) and software used in the study.

In the case of programming developed for dice detection and automatic registration, this will be available, for example, on GitHub.

Two different scanning stations are used in the experiments. It would be interesting to explain the replicability of the method for multiple stations:3,4,..n.

Lines 268 to 286: improve writing. First, introduce the study site and then the equipment used.

Please better justify the choice of the four different sites highlighting the characteristics considered in the selection of these sites.

Table 1: presents the density of points/average distance of points in the cloud. Is it possible to present the time required for data acquisition?

Results: Please compare the results obtained in your study with those obtained in the bibliographic references.

Line 318: Here we have a conclusion. Please reposition in the appropriate chapter.

Please check whether the conclusions adequately respond to the proposed objectives. Present proposals for future studies.

 

I end my review by congratulating them on the work done.

 

Respectfully,

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 4 Report

Dear authors,

For point clouds collected by laser scanning at close range (up to 12 m), an automated cube (pattern)-based registration method is presented that can provide registration accuracy comparable to the traditional sphere-based method. The proposed method is based on scanning a single piece of mold-like furniture (cubic target) or a large mold made of a lightweight material such as foam. The method first creates the coordinate systems using the lines between the intersections of the scanned faces to create the axes. Then, the non-orthogonal axes are corrected according to the projection geometry. This is followed by simulating three pairs of conjugate points along the corresponding axes for registration. The distances between conjugates are constrained to be one, so the main problem is to determine the relative position and orientation between the coordinate systems for the pattern space obtained from the two scans.

The theoretical part of the study and the experimental work are given in a simple way.

Tables and flowcharts are clear and understandable.

I had no problems with the fluency of the language.

Figure 7 can be rearranged.

I think the references are sufficient.

Author Response

Please see the attachment.

Author Response File: Author Response.docx