Siam-Sort: Multi-Target Tracking in Video SAR Based on Tracking by Detection and Siamese Network

Round 1

Reviewer 1 Report

The manuscript systematically discusses a multi-moving target detection and tracking method for video SAR images, including multi-target detection, data association and target state tracking. This is a carefully done study and the findings are of considerable interest. The scheme proposed in the manuscript has obvious reference value for the construction of SAR image processing system. However, the work would be excellent if the problem described below were modified and given some explanation.

1. The discussion of SAR data characteristics and video SAR imaging system characteristics should be added in this paper. These properties will be relevant to the algorithm design and further affect the performance and computational efficiency of the algorithm. The scheme given by the author of Section 2 only discusses the algorithm principle and implementation details from the perspective of image processing or visual computing. The technical framework adopted is inspired by typical algorithms in the field of computer vision and MTT, and some novel functional modules are designed accordingly, such as the anchors generation method in target detection.

2. The real-time performance of the algorithm is an important factor that needs to be paid attention to in the moving target tracking task, and this factor is directly related to the computing resource overhead involved in the deployment of the algorithm in practical application scenarios. Although the paper discusses the algorithm running speed indicator FPS in the experimental section, it is not specific enough for professionals in the SAR imaging processing community. It is suggested to discuss the real-time performance of algorithm deployment in combination with the imaging parameters of the video SAR system, such as imaging frame rate, spatial resolution, and image size, which will help the paper gain a wider audience in this field.

3. In Section 2.1.2, the function name "IOU" may be omitted in Equation 1, and the author needs to further check for possible clerical errors and make some minor modifications.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

In this manuscript, a multiple target tracking method based on TBD and the Siamese network are proposed. To improve the detection accuracy, the multi-scale Faster-RCNN is first proposed to detect shadows of moving targets. The authors claimed that dimension clusters are used to accelerate the convergence speed of the model in the training process as well as to obtain better network weights. Then, SiamNet is proposed for data association to reduce matching errors. At last, the Kalman filter is used to update tracking results. The author represents the experimental results on two real video SAR data demonstrating that the proposed method outperforms other state-of-art methods. And the ablation experiment verifies the effectiveness of multi-scale Faster-RCNN and SimaNet.

In this article, to achieve MTT in video SAR, a novel TBD-based method is proposed Siam-sort. Specifically, multi-scale Faster-RCNN is proposed to improve the detection accuracy for moving target shadows. Then, to further improve detection performance, the dimension clusters applied to give multi-scale Faster-RCNN a better prior in the training process. Also SiamNet, which used the similarities between features of target shadows. In two real video SAR data, Siam-sort acquired the best grades on all indicators, which demonstrated the proposed method outperformed MHT, SORT and Deep SORT. Additionally, the grades on all indicators were improved in the ablation experiment, which verified the effectiveness of multi-scale Faster-RCNN and SiamNet

The manuscript is clear, relevant for the field, and presented in a well-structured manner in general.

The number of cited studies is reasonable and most of them published recently.

 

The paper can be published as is.

 

Author Response

Dear reviewer，

Thank you so much for your reviewing! We deeply appreciate your recognition of our research work.