An Efficient Center-Based Method With Multilevel Auxiliary Supervision for Multiscale SAR Ship Detection

The problem of multi-scale ship detection in synthetic aperture radar (SAR) images has received much attention with the development of deep convolutional neural networks (DCNNs). However, existing DCNN-based multi-scale SAR ship detection methods often lead to time-consuming detection process due to the massive parameters therein. To address this issue, a lightweight center-based detector with the multi-level auxiliary supervision (MLAS) structure is proposed in this paper. First, an extremely lightweight backbone network is designed to improve the computation efficiency and extract SAR image features in a bottom-up manner. Then, a feature fusion network (FFN) containing three multi-scale feature fusion modules is introduced to combine semantic features with different levels. Finally, a novel MLAS-based framework is proposed to train our DCNN with multi-level auxiliary detection subnets. MLAS improves the performance of multi-scale ship detection benefiting from the guidance of multi-level attention. Experimental results on the open SAR image dataset SSDD show that our proposed detector achieves a similar average precision (AP) for the problem of multi-scale SAR ship detection but significantly reduces the computation burden of state-of-the-art methods. The required number of floating points of operations (FLOPs) of our method is only 21.70%, 19.30%, and 4.81% of those of CenterNet, YOLOv3, and RetinaNet, respectively, and the number of learnable weights in our method is only 0.68 million that is 5.63%, 1.10%, 2.98% of those of the aforementioned three existing methods, respectively.