Implementation of Real-Time Space Target Detection and Tracking Algorithm for Space-Based Surveillance

Round 1

Reviewer 1 Report

 - Authors use known techniques like the Hungarian method. It should be referred with proper citation and details of what solves in general such method and how it applies to the studied case. Same goes with the Kalman filet algorithm. Just in case the reader does not know well what these techniques are about.

 - Results are very nicely presented with promising outcomes. There is a nice comparison with SoTA proposals. Still, with 10^-4 false alarm rates, the diferencies do not look relevant. Is this just an impression due to the logarithmic scale? Please clarify.

English looks good.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The paper presents the proposal of a “Multi-stage Joint Detection and Tracking Model to solve the problem of space target detection and tracking under the deep space background and provided a hardware implementation of this model for space-based surveillance applications.” The paper's contributions and novelties are presented properly. The paper's readability is good, even with some minor typos and grammar errors. Similarly, the paper presentation is good, even with some minor issues regarding spaces, equations formalism, and figures size and fonts. Thus, the authors should consider proofreading. Additionally, the paper is technically sound. The proposed method is corroborated by experimental evaluations, and its performance is competitive when compared with other methods from the literature. As suggestions to improve the paper's quality, the authors should consider:

- Some of the presented block diagrams are not properly discussed. The authors should consider improving the description/discussion of the block diagrams in the Methodology Section.

- The author should consider including a more updated method for comparison in the evaluation presented in Figure 13.

- The authors should consider better addressing the implementation aspects of the proposed methods and better motivating the selection of the implementation parameters from the proposed method.

- More discussion regarding the FPGA and DSP implementations should be provided. The authors should motivate the selected methods/configuration setup for both implantations.

- More discussion regarding the observed results in Table 5 should be provided.

 

- Overall, more discussion should be included regarding the presented experimental/simulated results. The authors should consider including a discussion section in the current version of the paper.

The paper's readability is good, even with some minor typos and grammar errors. Similarly, the paper presentation is good, even with some minor issues regarding spaces, equations formalism, and figures size and fonts. Thus, the authors should consider proofreading.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

The manuscript proposed a multi-stage Joint Detection and Tracking Model (MJDTM) for space targets in optical image sequences, which combined a local contrast method and the Kalman filter to detect and track the potential targets and use differences in movement status to suppress the stellar targets. What is more, a heterogeneous embeded image processing system based on Field Programmable Gate Array (FPGA) and Digital Signal Processor (DSP) was established, and the proposed algorithm is efficiently implemented on this hardware platform. The research content is innovative and has practical value.

 

However, some improvement is still needed in this manuscript. In my opinion, the shortcomings of the manuscript are that the experiment analysis is not enough, the length of the manuscript is redundant, and the innovation points are not emphasized. The key parts of the proposed algorithm are not discussed in detail in the experiment section. Some of the major deficiencies are listed below:

1. In section 2, the authors reviewed the related works. The authors reviewed less works on the target trackers and the suppression of stellar targets than on the target detectors. It is suggested that the review of target tracking methods and stellar target suppression methods can be added more here.

2. In section 3, the authors introduced the target detection algorithm and the target tracking algorithm in detail. From my point of view, this manuscript can be reduced appropriately and focus on the innovative parts instead. For example, among the LFC operator/ the energy accumulation/ the adaptive threshold segmentation/ the centroid coordinates calculation and the constructed Kalman filter, which parts are the most important and innovative? and this innovation should be emphasized.

3. In section 3.2, formulas of (27), (28), (29) can create confusion in the reader's understanding. Only from two-dimensional coordinates cannot obtain the three-dimensional coordinates of a target. Here in the article, are the three-dimensional coordinates only the intermediate process? but ultimately to obtain the target’s two-dimensional coordinates of next moment based on the target attitude information?

4. How the attitude information of the satellite can be obtained and how the performance of the proposed stellar target suppression algorithm will change if there are errors in attitude information. This part of content can be added.

5. Based on the figure 1, the target feature (Target angle calculation mainly) is a part of output of the algorithm, but in the experiments section 5.2, target angles calculation test was mixed up with the camera distortion correction, and the manuscript did not give the calculation results of the target features based on the constructed dataset Seq 1/2/3/4.

6. In this manuscript, the necessity of using FPGA+DSP configuration for hardware system is not explained enough, for example, FPGA can also implement Kalman filter, whether it is possible to use only one kind of hardware. Also, the hardware system in this paper is designed to work on orbiting satellites, the test on hardware’s power consumption can be added to the experiment section of hardware system efficiency.

 

In addition, there are some minor problems with this manuscript:

1. Some abbreviations that appear for the first time are not given in their full form. Such as the ‘LCM’ in line 131, the ‘IR’ in line 134, ‘LVDS’ in line 523.

2. In the experiment, the stellar target suppression algorithm is not compared with other methods, is there any other existed simpler methods can be used comparing here?

3. The text in the picture in the manuscript should be bigger. such as the text in the left image of figure2.

4. The choice of LFC window size in the manuscript seems inconsistent in formula (2), (3) (s=4,7,10,13) and line 607 (LFC filter is set to 9*9).

5. Some hyperparameters appear in the proposed algorithm, and the authors do not explain how to select their values. such as the image down-sample rate and the ‘k1’.

6. The titles of subplots (b) and (c) in figure 7 are the same.

7. How are the space target image sequences simulated and acquired? Some details can be added.

8. The background of the Seq 4 data contains cloud, which is not consistent with the scenario of space target in the manuscript.

no comments

Author Response

Please see the attachment.

Author Response File: Author Response.docx