On the Use of Cameras for the Detection of Critical Events in Sensors-Based Emergency Alerting Systems
Round 1
Reviewer 1 Report
In the paper, the authors proposed a system of emergency detection and alerting based on scalar sensors and cameras as an extension to their work already proposed in [7]. In [7] only scalar sensors were considered. To accommodate the event detection through the cameras as sensors that visual monitor and identify critical events, the authors introduced to their original work a new type of event alert, the complex event alert. The authors as proof-of-concept implemented and showed experiences of the proposed system to validate the correct detection of critical events.
The overall paper is well presented and structure and describes well the issues and advantages of the adoption of cameras in emergency alerting systems. However, the reviewer major concerns are the following:
- The paper lacks in new contributions when compared with [7]. The paper results are limit to the demonstration of the system in operation which part was already presented in [7]. Moreover, the reviewer suggests the improvement of the results section. For example performance of the system results when occurs false alarm alerts of scalar sensors (e.g., faulty scalar sensors) or cameras sensors (e.g., bad image capture conditions).
- The authors mention the following “In order to assure robustness to the overall system, the Events Reports are constantly refreshed by the EDUs, resulting in a refreshing frequency for the alarms”. How the system performs, for example when an EDU alerts to a fire event but then gets damage, and due to that the EPU does not receive more event reports from this EDU.
Typo:
- line 345 - "0 < sl(a) <" should be replace by "0 < sl(a) < 1".
Author Response
Dear Reviewer
Thank you very much for your time and effort reviewing our article.
We carefully considered all of your comments and made our best to improve the quality of the article. We believe that the revised version is now better suitable for publication.
Comment 1:
"The paper lacks in new contributions when compared with [7]. The paper results are limit to the demonstration of the system in operation which part was already presented in [7]. Moreover, the reviewer suggests the improvement of the results section. For example performance of the system results when occurs false alarm alerts of scalar sensors (e.g., faulty scalar sensors) or cameras sensors (e.g., bad image capture conditions).”
Response:
Concerning this comment, since the paper is an extension of our previous work [7], we intended to provide additional substantial results to further contribute to the development of an effective emergency alerting system for urban areas. For that, we created a new proof-of-concept, implementing all functions of the new system and openly providing the codes in Github. Actually, the new proposed system treats the sensor units differently, demanding a new perception of the environment variables. In this sense, the results are organized in two parts:
1) A new implementation of all elements that compose the proposed emergency alerting system in the Python programming language. Although we “recycled" some elements of the work in [7], specially the communication code for the modules, substancial improvements were performed. All codes, as well as a reference implementation in the popular Raspberry Pi hardware platform, are freely available and ready to be used.
2) Concerning the amount and types of tests, we agree with you. We intended to perform more robust tests, but we had to handle with two big problems: a) the required cost for massive deployment of EDUs in the target city (Porto, Portugal) and b) the COVID-19 pandemic. This is why we were forced to perform most of the tests in indoor areas.
Particularly concerning your comment about the analyses of false alarms, they are indeed relevant! Analyses of the fault tolerance and dependability of the systems are very helpful, specially concerning critical systems. However, as we are most concerned in the description of a new emergency alerting system that is practical and ready to be used, some analyses of the fault resistance of emergency detection were put aside in the first moment. Nevertheless, we included a new subsection in the paper, “5.3. Failures and dependability of the system”, better describing all these issues. We believe that this new subsection contributed to the paper and we hope it addresses your comment.
Comment 2:
"The authors mention the following “In order to assure robustness to the overall system, the Events Reports are constantly refreshed by the EDUs, resulting in a refreshing frequency for the alarms”. How the system performs, for example when an EDU alerts to a fire event but then gets damage, and due to that the EPU does not receive more event reports from this EDU.”
Response:
This is also a very interesting question that demands special care.
This particular problem, and other failure conditions, were addressed in the new subsection “5.3. Failures and dependability of the system”.
Comment 3:
"line 345 - "0 < sl(a) <" should be replace by "0 < sl(a) < 1”."
Response:
We did not understand this suggestion, since the value of sl(a) will be between 0 and rmax. Actually, in the paper it was (100% of rmax), which may be confusing. We replaced that by 0 <= sl(a) <= rmax.
Thank you very much!
Reviewer 2 Report
The proposed approach in this paper for using cameras associated with scalar sensor looks very interesting and reliable moreover easy to implements and cost effective. However, I have some comments:
1) Please comment about the resolution and data processing of the used cameras since high resolution camera are not very cost effective and data require larger processing time and larger storage devices.
2) What is the advantage of your proposed system by combining scalar and visual compared to event trigged image sensor.
3) How about power consumption? Increasing sensors, both scalar and visual, will increase the overall power consumption dislike on such IoT application.
4) For urban use, how your proposed system is advantageous compared to existing and very sophisticated camera monitoring systems.
Author Response
Dear Reviewer
Thank you very much for your time and effort reviewing our article.
We carefully considered all of your comments and made our best to improve the quality of the article. We believe that the revised version is now better suitable for publication.
Comment 1:
"Please comment about the resolution and data processing of the used cameras since high resolution camera are not very cost effective and data require larger processing time and larger storage devices.
Response:
This is a very interesting suggestion. We further discussed these aspects in "Section 6: Practical issues when employing cameras for emergency alerting".
Comment 2:
"What is the advantage of your proposed system by combining scalar and visual compared to event trigged image sensor."
Response:
We believe that the advantage of such “union" is to achieve flexibility for the system, allowing easy definition of the critical events that will be detected by each EDU. In such way, the detection system will be as powerful as the employed visual computing algorithms, and as precise as the employed scalar sensors. However, event-triggered image sensors may also be used to detect events, as long as it is incorporated as a sensing unit of the EDU.
We made this clearer in the Introduction section.
Comment 3:
"How about power consumption? Increasing sensors, both scalar and visual, will increase the overall power consumption dislike on such IoT application."
Response:
We totally agree! We added a discussion about it in Section 6.
Comment 4:
"For urban use, how your proposed system is advantageous compared to existing and very sophisticated camera monitoring systems."
This is other very interesting remark! We discussed more about this in Section 6.
Thank you very much!
Reviewer 3 Report
Good aspects:
The ideas and contributions are clearly presented.
The paper is well written.
The solution was validated with an experimental platform. The sources of the application can be downloaded from Git Hub.
Weak aspects:
The paper is jut an improvement of the authors'work presented in reference [7]. Visual cameras were added to improve the detection of critical events. However, the paper does not address the algorithms to detect the critical events, it just presents how the visual cameras are integrated and how the corresponding events are treated by the proposed system.
The only relevant contribution of the paper is the integration of the visual cameras in the system for detecting events and sending reports, as well as the computation of the severity level when employing visual cameras. The experimental and results section does not contain a relevant validation of the computation algorithm. The validation scenarios are very simple. Some complex scenarios to show the solution's capacity to distinguish between the events (which one is critical and which is not) could help to validate the severity level computation.
There are no details about the costs of integrating the detection of events from visual sensors, especially the computation cost and power consumption costs.
The conclusion section could be improved.
Author Response
Dear Reviewer
Thank you very much for your time and effort reviewing our article.
We carefully considered all of your comments and made our best to improve the quality of the article. We believe that the revised version is now better suitable for publication.
Comment 1:
"The paper is just an improvement of the authors'work presented in reference [7]. Visual cameras were added to improve the detection of critical events. However, the paper does not address the algorithms to detect the critical events, it just presents how the visual cameras are integrated and how the corresponding events are treated by the proposed system.
The only relevant contribution of the paper is the integration of the visual cameras in the system for detecting events and sending reports, as well as the computation of the severity level when employing visual cameras. The experimental and results section does not contain a relevant validation of the computation algorithm. The validation scenarios are very simple. Some complex scenarios to show the solution's capacity to distinguish between the events (which one is critical and which is not) could help to validate the severity level computation”
Response:
The proposed system is indeed an extension of our work previously published in [7]. However, since the addition of the cameras to detect critical events is not a simple task, bringing new monitoring and efficiency issues to be handled, the paper intends to address most of the requirements and concerns when detecting events by the processing of visual data in conjunction with scalar data. In such way, a novel model for events classification, detection and alerting was proposed, which is highly flexible and adaptable to any urban context. In fact, the flexibility of the system to define target events and to inform the detection of such events through a multi-tier architecture is the main contribution of the paper. The performed experiment presented an example of fire detection using the opencv library, but any visual data processing can be used, since the architecture knows about it.
Cornering the experiments, we intended to perform more robust tests, but we had to handle with two big problems: a) the required cost for massive deployment of EDUs in the target city (Porto, Portugal) and b) the COVID-19 pandemic. This is why we were forced to perform most of the tests in indoor areas, limiting more extensive evaluations of the proposed approach. Nevertheless, the code provided in the Github directory was extensively tested and it is consistent for experimentations in bigger test scenarios. As future works, we plan to further validade the proposed approach.
Additionally, as an important discussion about the robustness and fault tolerance of the system, the new subsection “5.3. Failures and dependability of the system” was created, further discussing important experimental issues. Moreover, "Section 6: Practical issues when employing cameras for emergency alerting”, was revised and extended.
Comment 2:
"There are no details about the costs of integrating the detection of events from visual sensors, especially the computation cost and power consumption costs."
Response:
We agree that such details are relevant for the proposed approach. We extended “Section 6: Practical issues when employing cameras for emergency alerting” to better discuss such details.
Comment 3:
“The conclusion section could be improved."
Response:
The Conclusion section was improved, as suggested by the reviewer.
Thank you very much!
Reviewer 4 Report
This paper proposes an architecture based on visual sensors to detect critical events in an emergency alerting.
After detail review, I have some concerns which need to be addressed and are as follows:
(1) I think the paper needs a related work section after the introduction section or add literature in the introduction section. The authors should examine recent research.
(2) Authors should explain the problem more clearly in the introduction section.
(3) Remove unnecessary references from paper and add more related references, especially from 2019 to 2020.
(4) Paper contribution should be more clearly and explained in the introduction section.
(5) The paper has a type of error in many places. Check the form and correct all the typo errors from the paper.
(6) Some information is placed many times in different places. Please carefully check the paper and remove extra details.
(7) Revise the conclusion section with the proposed scheme's main contribution.
(8) Check references and try to edit some of them according to author names, issues number, and volume number.
(9) Add future direction with more details.
Author Response
Dear Reviewer
Thank you very much for your time and effort reviewing our article.
We carefully considered all of your comments and made our best to improve the quality of the article. We believe that the revised version is now better suitable for publication.
Comment 1:
"(1) I think the paper needs a related work section after the introduction section or add literature in the introduction section. The authors should examine recent research."
Response:
Thank you for your comment. However, we got a bit confused since the paper already had a Related Works section (Section 2). Anyway, we checked again the recent related works that are mentioned in the Introduction section and also discussed in the Related Works section. After careful analysis, we believe that the discussion of recent works is consistent and that the article as a whole has a good number of references (53). However, we further improved the Related Works section.
Comment 2:
"(2) Authors should explain the problem more clearly in the introduction section.”
Response:
We better explained the addressed problem in the Introduction section.
Comment 3:
"(3) Remove unnecessary references from paper and add more related references, especially from 2019 to 2020."
Response:
New references were added to the article, especially considering the period from 2019 to 2020. Unnecessary references were reviewed.
Comment 4:
"(4) Paper contribution should be more clearly and explained in the introduction section.”
Response:
We better described the contributions of the paper in the Introduction section.
Comment 5:
"(5) The paper has a type of error in many places. Check the form and correct all the typo errors from the paper.”
Response:
We performed a careful checking of the paper. The English writing was improved.
Comment 6:
"(6) Some information is placed many times in different places. Please carefully check the paper and remove extra details.”
Response:
Redundant information was removed from the paper, as suggested.
Comment 7:
"(7) Revise the conclusion section with the proposed scheme's main contribution.”
Response:
The conclusion section was revised and improved.
Comment 8:
"(8) Check references and try to edit some of them according to author names, issues number, and volume number.”
Response:
References were checked and eventual errors were corrected.
Comment 9:
"(9) Add future direction with more details.”
Response:
Future research directions were better discussed in Section 5, Section 6 and also in the Conclusion section.
Thank you very much!
Round 2
Reviewer 1 Report
In this revised version of the paper, the authors have addressed most of the reviewer's previous comments. They have improved the description of the sensor-based emergency alerting system and included some observation regarding the system failures conditions and practical implementation issues.
The reviewer believes that this work will help the researchers working in this topic. Thus, the reviewer recommends the paper for publication.
Reviewer 2 Report
The authors reviewed the paper as suggested and answered my all comments. The overall paper quality has been greatly improved compared to the first draft.
Reviewer 4 Report
I am satisfied with the revised version for the paper.
