Article Preview
Top1. Introduction
Radar is used widely in defense and public security fields, like early warning detection, sea surface, and air target monitoring. The radar echo has low observability, due to the presence of complex characteristics of the motion of the target and the presence of complex background environment, such as city, land, and ocean, etc. Thus, the moving target detection technology with low observability becomes the key factor that affected radar performance. If the target's energy is weak, it results in a low signal-to-noise/signal-to-clutter ratio (SNR/SCR) in both the frequency and time domain (Chen, et al., 2018). The SAR sensors perform military task, such as reconnaissance, wide-area battlefield surveillance, and intelligence gathering (Yan, et al., 2013). In addition to the monitoring of the environment (Bhambere, et al., 2021), terrain mapping, and the evaluation of the natural disaster (Cofini, et al., 2014, Rupapara, et al., 2020) damage for the application of civilians, they can obtain high-resolution images of targets and the ground scenes. In military activities, surveillance, and monitoring of urban traffic (Petruccelli, 2013), the ground moving vehicles' detection within the strong ground clutter is important. Most of the research on the SAR systems has drawn attention in terms of moving targets, such as estimating the motion parameter and detecting the moving target and location of the motion parameter (Xu, et al., 2018).
While considering the target detection in terms of clutter suppression, the observation space should effectively separate the clutter and the target. The observation space concerning the signal processing extends gradually to the sparse domain (Li, et al., 2018; Chen, et al., 2017a), frequency domain (Shui, et al., 2016), space-time domain (Melvin, 2000), and time-frequency domain (Chen, et al., 2017b) from the initial time domain. However, the dimension of the observation space is expanded further for the dynamic and complex background for providing more degrees of freedom (DoFs) in signal processing (Chen, et al., 2013). The radar detection is connected closely with the diverse and complex types of targets. Hence, detecting the target using radar signal processing has complexity in detecting the target and reducing overall detection performance (Chen, et al., 2018). Most of the methods for detecting the moving target include the optical flow method, frame difference method, background modeling method, and feature detection method. In the frame difference method, the target is extracted by considering the difference between two adjacent frames (Luan, et al., 2011). Although the frame difference method had good performance in real-time, it had low accuracy detection. Based on the image sequence's spatio-temporal gradient, the optical flow method estimated the motion field (Xin, et al., 2014). In the optical field method, the changes in the motion field are calculated for detecting the moving targets. The drawbacks of the optical flow method are the poor anti-noise ability, complexity in the calculation, and poor performance in real-time processing (Dong, et al., 2019).