A Visible Light 3D Positioning System for Underground Mines Based on Convolutional Neural Network Combining Inception Module and Attention Mechanism

Round 1

Reviewer 1 Report

This paper proposes a novel positioning system aiming for underground mines. The inception Module can extract information from received optical power at multiple scales, and the attention mechanism helps the model retrieve crucial feature information. By combining these two methods and selecting the optimal hyperparameters, the positioning system achieves an average error of 1.63cm and 11.12cm in the simulation environment and experimental scenario, respectively. 

 

1. The word of "ELM" that appears for the first time should have full spelling

2. The quality of Figures in the manuscript should be improved.

3. The expression of the sentence “ Typical 3Dlocalization methods typically rely on at least three LEDs and overlook the influence of wall reflections” is ugly, you may want to improve.

4. In Figure.3, the dimension of each layer should be expressed clearly

5. The output of the algorithm in table 1 is not the same as Figure.3

6. On page 8, why the units in online phase is 0.25m*0.25m*0.25m, but the offline phase is 0.2m*0.2m* 2m? You may want to explain how to choose these parameters and why they are different?

7. Lack of comparison with other method. It is better to compare the proposed method with other VLP scheme to better highlight your contribution.

8. The authors claim that “The proposed algorithmic model in this paper is characterized by its simplicity and ease of implementation.”. However, there are no experimental results to verify it, and according to the Table 5 on page 13, the training time of the proposed scheme is the largest one of 295.42s. So you may want to enhance the description to support your standpoint.

The writing should be substantially improved.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

This reviewer has the following two major concerns:

1.       What is the main motivation/reason for considering the underground mines as the target area? Technically, this does not seem to bring any specific challenges or requirements. The argument that the height changes, this can also be applied for almost all scenarios as the receiver moves. For example, the impact of receiver orientation has been investigated in many articles

2.       The main system motivation is not clearly justified. As I understand, the system has two components: the fingerprint database, which is created offline, and the neural network, which is trained offline, right? Both are also created or trained considering different PD heights, rights? What is then the advantage/requirement of having both of them? Please elaborate more on this.

No specific comments

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Issue 1.

This paper introduces a new system for accurately tracking the location of personnel in underground coal mines. The system uses a combination of a convolutional neural network (CNN), three-dimensional visible light positioning (VLP), and advanced techniques to overcome challenges such as fluctuating heights, tilted photodetectors, and wall reflections. By extracting multi-scale features and assigning varying weights to different channel features, the system achieves a high level of accuracy. Simulation and experimental results show that the system achieves centimeter-level positioning accuracy, meeting the requirements for personnel tracking in coal mines. Accurate personnel positioning in underground coal mines is crucial for ensuring safety and effective management of workers' locations and movements.

 

Issue 2.

An introduction section must be enlarged. Moreover, an enhanced literature review is required.

In the introduction section of the mentioned research paper, the following issues could be raised:

- The limitations and challenges of existing positioning systems in coal mines, such as accuracy, robustness in dynamic environments, and the impact of factors like fluctuating heights, tilted photodetectors, and wall reflections.

- The need for advanced technologies, such as convolutional neural networks (CNNs) and visible light positioning (VLP), to address the limitations of current systems and achieve higher accuracy and reliability.

- Cases of successful implementation of machine learning and deep learning in mining.

- The potential consequences of inaccurate personnel positioning, including increased risk of accidents, slower emergency response times, and difficulties in coordinating and optimizing mining operations.

 

Issue 3.

In case to add more information (literature review according to Issue 2, point 3) please consider the suggested research in your paper when enhancing the introduction section. I believe they are worth considering in your paper.

Jinqiang, W.; Basnet, P.; Mahtab, S. Review of machine learning and deep learning application in mine microseismic event classification. Min. Miner. Depos., 2021, 15, 19-26. https://doi.org/10.33271/mining15.01.019

Shen, Z.; Deifalla, A.F.; Kamiński, P.; Dyczko, A. Compressive Strength Evaluation of Ultra-High-Strength Concrete by Machine Learning. Materials 2022, 15, 3523. https://doi.org/10.3390/ma15103523

More recent references are more than welcome.

 

Issue 4.

Methodology (indicate lines where raised issues were considered):

- How does the proposed convolutional neural network (CNN) three-dimensional (3D) visible light positioning (VLP) system utilize the Inception-v2 module?

- What are the specific challenges addressed by the system in relation to fluctuating heights, tilted photodetectors, and wall reflections?

- How does the system extract multi-scale features from the optical power data acquired by the photodetectors using the Inception module?

- What is the role of the efficient channel attention mechanism in the system, and how does it contribute to the accuracy of personnel positioning?

- Can you provide more details on the simulation and experimental results, such as the specific methods used for evaluation, the size of the localization environment, and the positioning error achieved in both scenarios?

 

Issue 5.

 

After careful evaluation, I must acknowledge the exceptional quality of the study conducted, and I am highly inclined to recommend your paper for publication. However, I suggest conducting a meticulous revision to ensure that all aspects of the research are refined and polished before final submission.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Thanks for your revision.

It's OK.

Author Response

We are honored to receive your recognition and thank you very much for your instructive comments.

 

Reviewer 2 Report

I first would like to thank the authors for revising the article following my comments. However, the authors did not provide sufficient answers and the main points still not clear:

1. my first comment was not about the motivation for using the visible light for positioning, but the motivation for considering the mining as the target indoor environment. In other words, what is (are) the unique technical challenge(s) in the mining environment which is(are) not in other environments?

2. I still do not see the main role of the neural network in this system. The fingerprint database is obtained offline by collecting the received signal level at different positions, either in 2D or 3D planes. In other words, for each position in 2D or 3D plane, there is a corresponding signal level saved in the database. What is then the role of the neural network? What features to be extracted using such a network, as all information already stored in the database? The authors should make this point clear. 

No specific comments

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Very good revision.

I suppose the paper is ready for publication.

Author Response

It is a great honor to have you recognize this work, which will be my motivation in my next work！

Round 3

Reviewer 2 Report

Thanks for your responses. No further comments

No specific comments