24 December 2021 in IEEE Transactions on Geoscience and Remote Sensing
Aiming at the technical difficulty of marine radar to detect small targets embedded in the sea clutter, this article proposed a three-feature fusion detection method based on singular spectrum analysis. First, considering that the number of coherent pulses used by radar in scanning mode is usually small (64 or less), this method combines the application of radar historical scan data and current frame data, transfers the feature extraction method from intraframe to interframe, and extracts three features that consist of cumulative major singular value (CMSV), linear degree of second singular vector (LDSSV), and linear degree of third singular vector (LDTSV) from singular space of the cell under test (CUT). Second, in view of the unideal distribution of sea clutter samples, a 3-D concave hull learning algorithm based on the geometry shape of sea clutter samples under the framework of anomaly detection is developed by improving the original convex hull algorithm, and target detection is realized in feature space using this algorithm. Under the same parameter condition, the measured CSIR data verify the two following points: first, the performance of detector using concave hull learning algorithm is better than that of convex hull learning algorithm; second, the detection performance of the proposed detector is obviously better than that of tri-time–frequency (TF)-feature detector, trifeature-based detector, consistency factor detector, and fractal-based detector.
4 July 2022 in IEEE Transactions on Geoscience and Remote Sensing
Under the framework of feature-based detection of small targets on sea surface, existing feature extraction methods only use the echo data of current frame while ignoring the influence of historical echo data. Nevertheless, due to the non-stationarity of sea clutter, it may lead to unstable extraction of detection features, and then affect detection performance. To solve this problem, this paper designs a feature extraction method based on a priori information for small target detection. It firstly obtains a priori information from historical echo data by kernel density estimation (KDE) method. Then, the corresponding feature estimation method is utilized to obtain improved feature according to the relationship between current frame data and a priori information. Finally, the feature information of current frame is integrated into a priori information to prepare next feature extraction. Measured data are utilized to verify the performance of proposed method and the results reveal that, this method can effectively improve detection performance especially when sea clutter and target echo have good separability. In addition, the complexity of algorithm is analyzed to prove that proposed method has certain application potential.
1 January 2025 in IEEE Transactions on Instrumentation and Measurement
11 February 2025 in IEEE Transactions on Aerospace and Electronic Systems
Currently, reinforcement learning (RL) has been applied for the multi-target detection task of MIMO radar. However, the existing methods still have two shortcomings: 1) the detection performance on weak targets is insufficient, and 2) the solving time of beam optimization scheme (BOS) is long. For the first issue, this paper first proposes a partially time-varying statistic vector, which can more robustly distinguish the target cell from clutter cell during the “action” step of RL. Secondly, considering the scenario where both strong and weak targets exist, a BOS with strong target limitation is proposed, which aims to limit the power gain of beampattern on strong targets. Thirdly, considering the scenario with all weak targets, the search mode of radar is given, and its corresponding BOS is designed. The aim is to search for possible weak targets by raising the power gain within a part of observation area. For the second issue, by introducing multi-rank beamformer, the existing and proposed BOSs are transformed into corresponding convex minimax constraint optimization problems, and their closed-form solutions are derived. Accordingly, a fast solution method is proposed. By combining the above improvements with the Markov decision process model of MIMO radar, as well as the quasi E-greedy policy and the optimized reward mechanism in existing methods, the proposed method is given. Sufficient experiments verify that the proposed improvements are valid, and the proposed method owns better detection performance compared with its competitors.
12 January 2024 in IEEE Geoscience and Remote Sensing Letters
Under the background of slow small target detection on sea surface, it is a serious problem that the detection performance of existing feature detection methods decreases when the number of coherent pulses is less. In this letter, we first model the slow small target detection on sea surface as a contextual bandit (CB) problem. On this basis, we propose an autonomous feature detection method by modifying the classical feature detection process. The method can autonomously choose the detectors with better performance under the current sea scene from the constructed 18 feature detectors and obtain the fine detection performance by fusing their detection results. The performance superiority and the real-time capability of the proposed method are verified by the experiments on seven CSIR datasets.
27 August 2024 in IEEE Sensors Journal
Currently, the contextual bandit (CB) has been applied in the field of slow-moving small target detection on sea surface, to solve the performance decline problem of feature detection method under less coherent pulse number. However, the existing method only gives a brief description on its implementation details, and also has two shortcomings: 1) the algorithms used for estimating target location lacks theoretical support, and 2) the adopted policy is idealized. For the first issue, we assess the rationality and robustness of the initial target location estimation method, the online update algorithm for estimated target location and the suppression algorithm for successive false alarms, both theoretically and experimentally. Then, we summarize the above three parts and propose a real-time target location estimation method. For the second issue, by modeling a nonlinear relationship between the expected reward of action and its observed context, the kernelised upper confidence bound (KernelUCB) policy is introduced. Subsequently, a k-nearest neighbor based approach is proposed to reduce its algorithm complexity. Further, we propose designing an alternative action, which is enabled once main action chosen by policy be invalid, thereby improving the performance of policy. By combining the above improvements with the CB process modeled from classical feature detection process, proposed method is obtained, which can realize the autonomous optimized selection for alternative feature detectors on each frame. The sufficient experiments demonstrate that proposed method can integrate the performance advantages of various feature detectors, while meets the real-time detection requirement.