Traffic Monitoring from the Perspective of an Unmanned Aerial Vehicle
Round 1
Reviewer 1 Report
The manuscript introduces a traffic monitoring system based on the videos recorded by unmanned aerial vehicles. The system leverages an object detection algorithm (i.e., YOLO) and designs a trajectory tracking mechanism for better traffic monitoring. Finally, the implementation and the evaluation are given to readers.
The organization of the manuscript is appropriate. However, the main concern of the manuscript is the lack of a clear problem statement and the description of innovation and contribution. It is suggested that the authors can elaborate on the problem statement, the proposed solution, and the improvement in more detail.
The comments and suggestions are listed as the following.
1. Unmanned aerial vehicle is claimed to be involved in the study. However, there is little information about the UAV. For instance, UAVs can move around and record videos at different locations. The impact of the mobility should be described. In addition, different aerial situations might affect the UAV and thus affect the quality of videos (e.g., camera shaking effects). Thus, misdetection, misclassification, and mistracking will happen during the analysis phase. Since the title of the manuscript is "Traffic monitoring from the perspective of unmanned aerial vehicle", it is strongly suggested that the authors should elaborate on the characteristics and potential challenges brought by the UAV. Thus, readers of the journal can get more insights into the contribution of the manuscript.
2. According to line 22, the manuscript states that "The research described in this paper proposes and experimentally implements a web-based service for environment monitoring." The effort of the design and implementation is appreciated. However, the innovation should be addressed in more detail. In addition, why "environment monitoring"?
3. According to the summary of the contributions of the manuscript, the manuscript states that "Assessment of the UAV's potential to monitor the temporal state of the traffic in multiple locations and determine the suitable scenarios". The assessment should be given in the manuscript. For instance, the advantages and disadvantages of using UAVs in traffic monitoring can be described and compared with cameras. In addition, more scenarios should be given and discussed (e.g., weather condition, traffic condition, time, and so on).
4. According to the summary of the contributions of the manuscript, the manuscript states that "Verify and optimize the algorithms for video analysis following sample scenarios". If optimization is claimed to be the contribution of the manuscript, a comparison with the literature should be given. Thus, readers can get insight into the improvement of the proposed algorithm. Moreover, a comprehensive literature review is necessary.
5. In Figure 1, the images taken by UAVs are presented. It is expected that videos with different views by UAVs will be the input of the proposed system. Furthermore, the system should be able to analyze those videos. However, in the evaluation part, only two sample videos are used. The evaluation might be insufficient.
6. According to line 240, the manuscript states that "we determine several enhancements to reduce the hardware requirements and increase the chance to get the most reliable data." First of all, the enhancements should be quantified and described. Secondly, based on the architecture, the computation is performed on the server-side, and not on the UAV in real-time. Thus, the reduction of hardware requirements might not be necessary. Finally, the method to get reliable data should be elaborated in more detail.
7. An architecture diagram is depicted in Figure 5. However, this seems a general one. Please specify the improvement or contribution compared with previous studies.
8. Section 3.5 describes the results of video analysis. It is suggested that the authors can elaborate on what analysis the proposed system can do (e.g., accident, violation, traffic jam, and so on).
9. Section 4.2 describes the evaluation of the object tracking algorithm. It is suggested that the authors can provide quantitative data and comparison with previous studies. Thus, readers can get more insight into the improvement and contribution of the proposed algorithm.
10. In Section 5 (Conclusion), the manuscript states that "Unfavorable wind conditions lead to a significant camera movement, which may distort results". This is a challenge. However, related discussions and solutions cannot be found in the manuscript.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
dear authors:
the paper is very good. However, i have two comment
1- YOLOv4 model needs to be present in detail with all mathematical equtions.
2- explain the data set in more detail.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
The use of UAVs for operational monitoring is an important task. The authors consider the effectiveness of their web-based solution in good sunny weather, and the question arises how much will the effectiveness of the web-based solution decrease in rain, fog or at night? This should be spelled out in the manuscript.
Is it possible to further expand the functionality of the proposed web-based solution, for example, can it be adapted for monitoring using a drones at the airport (proposed in: 10.1007/s12198-020-00225-z), or for monitoring using drones at high-speed rail (proposed in: 10.1504/IJCIS.2019.103015), i.e. there it is necessary to control the traffic of vehicles with significantly different dimensions.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The revision of the manuscript is appreciated. The authors also address several comments and suggestions in the previous review. According to the current manuscript, some comments are listed as the following. It is suggested that the major research question, solution, and contribution can be described in more detail for readers.
Contribution 1: Assessment of the UAV's potential to monitor the temporal state of the traffic in multiple locations and determine the suitable scenarios.
Comment: It is suggested that the authors can provide a table for the comparison (i.e., advantages and disadvantages) of different monitoring mechanisms (e.g., UAV, static camera, and camera mounted on car). Also, the challenges of those mechanisms can be described for readers. Thus, readers can get valuable insights when they design and implement traffic monitoring systems.
Comment: As stated in the manuscript, different flight altitude and weather conditions will affect the videos recored by UAV. Can the proposed system deal with this issue? Any video analysis and the following evaluation about different conditions (scenarios)?
Contribution 2: Design and implement the experimental software solution for the detection of vehicles and their direction.
Comment: The design and implementation of the software solution by Open Source Software is appreciated. However, the architecture seems to be a general one. Will the architecture change when the input video is recored by the static camera? If not, it is a general solution for video analysis. In other words, what are the major concerns about the architecture/algorithm design when analyzing UAVs' videos? Does the proposed system deal with those concerns? Any improvement (e.g., the flexibility of setting ROI?) of the proposed system compared to literature? Thus, readers can get more insights into the contribution brought by the manuscript.
Contribution 3: Verify the algorithms for video analysis following sample scenarios.
Comment: Based on the statement, the manuscript verifies several algorithms for video analysis. Does the manuscript propose new algorithm for object detection or object tracking?
Comment: Based on the manuscript, it seems the authors have a new algorithm for object tracking. If it is true, please elaborate on the algorithm (e.g., math formula or pseudo code) in more detail. If existing object tracking method is used, please specify the reference. In addition, the evaluation can be given (e.g., the proposed object tracking algorithm and Deep SORT).
Overall, the organization of the revised manuscript is appropriate. If the authors can specify the problem statement, problem solution, and contribution in more detail, the readers of the journal can get benefit from the manuscript.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
I accept the revised manuscript
Author Response
Thank you for the comments. The conclusion was revised. Also, other sections of the paper were revised to provide a better reading experience for the readers.
Round 3
Reviewer 1 Report
The revision of the manuscript is highly appreciated. The manuscript is recommended for publication.
