On the remote sensing images of marine synthetic aperture radar (SAR), ship targets often occupy only a small part of an image, and the rest are all sea and coastal backgrounds. Existing neural networks based on SAR ship detection often directly detect an entire SAR image, which consumes a large amount of computing resources. In this paper, a new marine SAR ship detection network, called multilevel background removing–you only look once X (MLBR-YOLOX), is proposed. First, a new plug-and-play module, called standalone spatial patch detector (SSPD), is proposed to pre-detect the position of ships and filter out most of the sea backgrounds at the source image level. Second, a deep spatial feature detector (DSFD) is presented for detecting deep semantic features of the output of the backbone module to further reduce the computational cost of the neck and head modules. Finally, the original YOLOX network is adopted to locate and classify the pre-detected results. Experimental results on the SAR ship detection dataset (SSDD) and the high-resolution SAR images dataset (HRSID) indicate that the detection performance of MLBR-YOLOX is close to that of YOLOX, but the computational complexity is merely 23.97% and 12.50% of YOLOX's, respectively. Moreover, the experiment conducted on a large-scene Sentinel-1 SAR image illustrates that the proposed network has good migration application capability.
21 July 2014 in IEEE Transactions on Aerospace and Electronic Systems
An adaptive digital beamforming technique with compressed sensing (CS) for sparse receiving arrays is proposed. Because of the angle sparseness of arriving signals, CS theory can be adopted to sample receiving signals. Then, receiving signals from absent elements on the antenna aperture can be reconstructed by using CS method. Adaptive digital beamforming algorithms are adopted to form antenna beam, whose main lobe is steered to the desired direction and nulls steered to the directions of interference.
4 October 2018 in IEEE Antennas and Wireless Propagation Letters
Motivated by beam distortions in multiple interferences environment, we propose a novel combined constrained method for adaptive sum and difference beamforming in monopulse angle estimation. By optimizing the adaptive weights of sum and difference beams jointly, the restriction on the degree of freedom to determine the adaptive sum or difference weight is evenly distributed to all weights. The constraints of array gain and monopulse ratio curve are utilized to avoid beam distortion while nulling the interferences. Simulations of the proposed algorithm have been provided, which show a significant improvement in beamforming and angle estimation performance compared with the conventional constrained adaptive monopulse algorithm.
In this letter, a new array synthesis method called weighted alternating projection and proximal splitting is proposed, which combines the alternating projection (AP) and constrained weighted least mean squares (CWLMS). Under the framework of AP, we first study the projectors over the pattern and excitation constraints in detail. Especially for the projector over excitation constraints, it is designed as the solution to a CWLMS optimization. Then an effective proximal splitting method is developed to solve the CWLMS optimization using proximity operator theory. To demonstrate the performance of proposed method, several planar synthesis examples with embedded element patterns are presented and the results show that the proposed method has better performance.
21 July 2016 in Journal of Systems Engineering and Electronics
By applying phase-only technique in array antenna pattern synthesis, antenna arrays can form desired patterns with the use of phase shifters only. A novel phase-only pattern synthesis algorithm is proposed for the passive phased array seeker. This algorithm synthesizes the main beam of the antenna pattern through least-squares approximation, thus minimizing the errors between the actual and the desired main beams. The synthesis problem can be solved by applying gradient-descent optimization. The item for suppressing side lobes is added to the above synthesis problem. To obtain a side lobe level as low as possible, the algorithm assigns different weights to different directions in the side lobe region. The algorithm is run repeatedly and the weights are adjusted adaptively according to the normalized power in the side lobe directions. Detailed examples are presented to demonstrate the accuracy and effectiveness of the proposed approach.