Ship instance segmentation in synthetic aperture radar (SAR) images is a hard and challenging task, which not only locates ships but also obtains their shapes with pixel-level masks. However, in ocean SAR images, because of the consistent reflective intensities of ships, the appearances of different ships are similar, thus making it far too difficult to distinguish ships when they are in densely packed groups. Especially when ships have incline directions and large aspect ratios, the horizontal bounding boxes (HB-Boxes) used by all the instance-segmentation networks that we know so far inevitably contain redundant backgrounds, docks, and even other ships, which mislead the following segmentation. To solve this problem, a novel ship instance-segmentation network, called SRNet, is proposed with rotated bounding boxes (RB-Boxes), which are taken as the foundation of segmentation. Along the directions of ships, the RB-Boxes can surround the ships tightly, but a minor deviation will corrupt the integrity of the ships’ masks. To improve the performance of the RB-Boxes, a dual feature alignment module (DAM) was designed to obtain the representative features with the direction and shape information of ships. On account of the difference between the classification task and regression task, two different sampling location calculation strategies were used in two convolutional kernels of the DAM, making these locations distributed dynamically on the ships’ bodies and along the ships’ boundaries. Moreover, to improve the effectiveness of training, a new adaptive Intersection-over-Union threshold (AIoU) was proposed based on the aspect-ratio information of ships to raise positive samples. To obtain the masks in the RB-Boxes, a new Mask-segmentation Head (MaskHead) with the twice sampling processes was explored. In experiments to evaluate the RB-Boxes, the accuracy of the RB-Boxes output from the Detection Head (DetHead) of SRNet outperformed eight rotated object-detection networks. In experiments to evaluate the final segmentation masks, compared with several classic and state-of-the-art instance-segmentation networks, our proposed SRNet achieved more accurate ship instance masks in SAR images. The ablation studies demonstrated the effectiveness of the DAM in the SRNet and the AIoU for our network training.

For ships in synthetic aperture radar (SAR) images, which have various aspect ratios and incline angles, rotated regions of interest (ROIs) are used in instance segmentation recently as compact segmentation patches. Although the segmentation accuracy of ships has been improved with rotated ROIs, their contour details are unsatisfactory. To refine the contour details, a dual-scale-feature-fusion segmentation network, called DRSNet, is presented in this letter around rotated ROIs. To upgrade performance of ROI detection, the Gaussian-distribution-based ATSS (Ga-ATSS) is proposed to raise positive samples according to shape information of ships for training of the bounding-box detection module (DetHead). To enhance segmentation of contour details, a dual-scale cross attention (DS Cross Attention) is designed to extract accurate detail features from low-scale features and add them into high-scale semantic features in the rotated ROIs. The experimental results reveal that the DRSNet can achieve average precisions (AP) of 67.5% on the SAR ship detection dataset (SSDD) and 68.1% on the instance rotated ship detection dataset (Instance-RSDD), which outperforms the comparison networks. And ablation study shows that the Ga-ATSS can improve the ship detection performance, especially Recall rises by 1.3%.

How to effectively identify vehicles in SAR (synthetic aperture radar) images is a challenging problem in the SAR image understanding community. Compared with the bounding-box-based localization by object detection networks, object segmentation could provide the complete description of vehicles, which not only detects vehicles’ positions but also obtains their contours. However, vehicle segmentation in SAR images is a hard task due to speckle noise. To address this problem, this letter proposes a focal-pyramid-based network for vehicle segmentation in SAR images, called FVSNet, consisting of a focal pyramid module and a segmentation module. The focal pyramid module is built for extracting the multi-scale semantic feature maps from inputted SAR images, where the channel weights of the feature maps from backbone are automatically adjusted via a channel attention and the ambiguous features derived from speckle noise are eliminated via a designed mask-based cross attention. The segmentation module is designed to enlarge the feature maps outputted from the focal pyramid module with higher-resolution feature maps to obtain vehicle segmentation results. In addition, considering that there exists no open-source manually annotated dataset for handling the SAR vehicle segmentation task, we build and release a SAR vehicle-segmentation dataset based on the public MSTAR dataset. Experimental results on this dataset demonstrate the effectiveness of the proposed network in comparison with eight existing networks.

In the field of autonomous driving and robotics, 3D object detection is a difficult, but important task. To improve the accuracy of detection, LiDAR, which collects the 3D point cloud of a scene, is updated constantly. But the density of the collected 3D points is low, and its distribution is unbalanced in the scene, which influences the accuracy of 3D object detectors in regards to object location and identification. Although corresponding high-resolution scene images from cameras can be used as supplemental information, poor fusion strategies can result in decreased accuracy compared with that of LiDAR-point-only detectors. Thus, to improve the detection performance for the classification, localization, and even boundary location of 3D objects, a two-stage detector with density-and-sparsity feature aggregation, called DASANet, is proposed in this paper. In the first stage, dense pseudo point clouds are generated with images from cameras and are used to obtain the initial proposals. In the second stage, two novel feature aggregation modules are designed to fuse LiDAR point information and pseudo point information, which refines the semantic and detailed representation of the feature maps. To supplement the semantic information of the highest-scale LiDAR features for object localization and classification, a triple differential information supplement (TDIS) module is presented to extract the LiDAR-pseudo differential features and enhance them in spatial, channel, and global dimensions. To increase the detailed information of the LiDAR features for object boundary location, a Siamese three-dimension coordinate attention (STCA) module is presented to extract stable LiDAR and pseudo point cloud features with a Siamese encoder and fuse these features using a three-dimension coordinate attention. Experiments using the KITTI Vision Benchmark Suite demonstrate the improved performance of our DASANet in regards to the localization and boundary location of objects. The ablation studies demonstrate the effectiveness of the TDIS and the STCA modules.