Object Tracking Based on Optical Flow Reconstruction of Motion-Group Parameters
, George Petkov 1, Sylvia Ilieva 1, Alexander Petkov 2 and Stiliyan Kalitzin 3,4,*Round 1
Reviewer 1 Report (Previous Reviewer 1)
Comments and Suggestions for AuthorsThe paper presented an optical flow based moving object tracker.
1. Please include in the manuscript what are the differences between the work presented in this paper and the authors’ previous work (example GLORIA algorithm).
2. Please include in the manuscript qualitative results for the proposed technique and the baseline method (GLORIA) on LaSOT dataset and support these results with a discussion.
3. Please also report in the paper the tracking speed of the proposed technique in frames per seconds (FPS) when LaSOT dataset is used.
Comments on the Quality of English LanguagePaper title still a question, how "remote" is relevant to this research work without evidence in the experimental results. Therefore, the paper title and the while manuscript need revision to correct this terminology.
Author Response
Please see the file attached
Author Response File: 
Author Response.pdf
Reviewer 2 Report (Previous Reviewer 2)
Comments and Suggestions for AuthorsAuthors have addressed and incorporated all the questions and comments into the revised version. This paper can be accepted.
Author Response
Please see the file attached
Author Response File: Author Response.pdf
Reviewer 3 Report (Previous Reviewer 3)
Comments and Suggestions for AuthorsAuthors are presenting an algorithm for object tracking based on optical flow reconstruction. Object detection is obviosly much investigated field due to many possible applications and challenges. I have following comments on this work:
- Introduction does not correspond to the rest of the paper. Although it is focused on medical applications and particulary epileptic patients, in the rest of the paper (examples and tests) we can’t find anything related.
- There is no significant scientific contribution in paper since the core of the algorithm is not new and only change is related to adaptable ROI..
- Diagram presented in Figure 2 is quite simple and seems strange (far from any existing diagram presentation standard).
- Results don’t bring any relevant comparison to some larger dataset. Presented comparison with a sequence from LaSOT dataset is not nearly enough.
- Where is the proof for statement presented in lines 354-357?
- Also, some of the presented figures in the Results section are quite confusing and not informative at all. For exaple, what can a reader conclude from Figure 17?
- Conclusion should be better related to previous text. In line 495 authors finally mentioned detection of patient movements. However, this is first time after the introduction that this topic occur. Where are tests for this?
- Structure of the paper changed.
Comments on the Quality of English Language
Language should be significantly improved. Senteces are often to long.
Author Response
Please see the file attached
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report (Previous Reviewer 1)
Comments and Suggestions for AuthorsThe paper presented an optical flow based moving object tracker.
My main comment in this round is as follows: Section 3 requires reorganization and to have clearer sub-sections titles to make it easier to read and follow. For instance, the transitions from one experiment to another is not clear, i.e., shifting from results on simulated data to those on sample test videos with manually selected ROI and evaluating the method on LaSOT dataset require clearer subsections. Therefore, adding more subsections titles / paragraph headers will make the flow of the ideas in this section easier to understand.
Comments on the Quality of English LanguageN/A
Author Response
We appreciate this helpful suggestion by the Reviewer; indeed the “container” with results was not clearly organized. In the revised manuscript, we have added sub-titles delineating each experiment and shortly indicating the objective.
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
Comments and Suggestions for Authors
The paper presented an optical flow based moving object tracker. The study and the simulation results are interesting, while the presentation of the paper and the experimental results are not aligned with the recent advancement in the single object tracking field.
1. The paper needs to clarify the novelty and supporting the claims with experimental results. For instance, in the abstract the paper mentioned “The method successfully applies to multi-spectral data,...” while there are no experimental results in the paper that can justify this claim. The same problem with the last sentence of the abstract “The proposed optical flow reconstruction provides accurate, robust, and faster results compared to pixel-level algorithms and segmentation-based approaches “, the paper does not include any comparison with other methods (please read the next comment). The term “Remote object tracking” in the title and “remote tracking” in several places in the paper, is not supported by any experimental results, does “remote” refer to remote sensing as discussed in Section 4 without any experimental evidence, or remote functionality as mentioned in Section 4 that looks like a communication problem rather than a video analysis one.
2. The experimental results section (Section 3) includes interesting simulation results, while the paper is lacking quantitative and qualitative comparisons with recent state-of-the-arts single object trackers on recent publicly available visual object tracking datasets [R1-R3]. The paper also does not include comparison with the baseline method (GLORIA). Therefore, Section 3 requires additional experiments and discussions.
3. The introduction section lacks elaboration and citation for the recent object tracking methods and survey papers. For instance, the paragraph in lines 34-41 should include citation for the mentioned object tracking methods / approaches.
4. In Figure 14, it's not clear where's the target ROI?
5. Equation 6, please check “L_y = Y_4 - Y_2 = Y_3 - Y_1”?
6. In Figure 2, the block named “ROI elements update...” is repeated twice, while the loop could be directed to the first time this block is mentioned. Please revise the figure.
[R1] Fei Chen, Xiaodong Wang, Yunxiang Zhao, Shaohe Lv, and Xin Niu. "Visual object tracking: A survey." Computer Vision and Image Understanding 222 (2022): 103508.
[R2] Xi Li, Weiming Hu, Chunhua Shen, Zhongfei Zhang, Anthony Dick, and Anton Van Den Hengel. "A survey of appearance models in visual object tracking." ACM transactions on Intelligent Systems and Technology (TIST) 4, no. 4 (2013): 1-48.
[R3] Milan Ondrašovič, and Peter Tarábek. "Siamese visual object tracking: A survey." IEEE Access 9 (2021): 110149-110172.
Comments on the Quality of English Language
a. Sentence at Lines 44-45 are hard to read. Please revise.
b. Line 253, “Linear correlation between is much lower…” Please revise.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThis paper proposes a new optical flow method for object tracking. The need for such a method is not comprehensively discussed. The study seems interesting, but the reviewer has the following questions, comments, and concerns that need to be addressed in the revised version of the paper.
1. Fig4.a) and b) are not properly discussed. Please discuss each subfigure in detail.
2. Why in Fig.4b) there is too much difference in the true movement value and calculated value? Please also quantify the estimation error with some proper metric.
3. Fig 5.b) is not properly explained. Please also explain why there is such a big difference in input dilation and calculated dilation.
4. Please perform an exhaustive comparison with the state-of-the-art object tracking methods such as YOLO, Simple Online and Realtime Tracking (SORT), DeepSORT (https://arxiv.org/pdf/1703.07402.pdf), FairMOT (https://arxiv.org/pdf/2004.01888v6.pdf), phase based tracking (https://ieeexplore.ieee.org/document/8737366 ) etc.
5. Also validate your algorithm of the SOTA object tracking datasets (https://paperswithcode.com/datasets?task=object-tracking&page=1 )
6. Please also explain the pattern of Fig.15.
7. Is it possible to track the object if the camera is moving? Please consider looking into this paper for reference (https://onlinelibrary.wiley.com/doi/full/10.1002/stc.2713)
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsAuthors present an object tracking procedure based on optical flow reconstruction. This is generally a popular and relevant topic but this paper has some serious flaws.
Abstract is to general, it should provide more specific info on proposed method. Introduction is very superficial and doesn't provide introspection of research motivation, current trends and newer literature. Generally, references are much outdated. Introduction is poorly structured and even some language sty should be improved (for example, lines 44-45). Stated objective (tracking a single object that moves in camera FOV) is naive. Overall, reader doesn't receive any info on scientific contribution of this research (mainly because it lacks novelty(.
At some point it seemed that this paper could have future if it is rewritten with focus on evaluation of (and proposing a new) method tracking of epilepsy patients. Unfortunately, in later sections there was no further mention on this particular area of interest.
Presented method is based on already published GLORIA algorithm so there is no real novelty here.
Presentation of method should be improved. For instance, diagram in fig. 2 should be optimized since it is quite redundant and also doesn't provide any additional info (frames 1 and 2 can also be presented as frames i-1 and i so there are some unnecessary blocks in the image).
Results - missing a real evaluation of object tracking on some public video database. Also, comparison with the state-of-the-art methods is missing.
Comments on the Quality of English LanguageSome parts should be improved.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
