Anchor-Free Arbitrary-Oriented Object Detector Using Box Boundary-Aware Vectors
Donghang Yu,Qing Xu,Haitao Guo,Junfeng Xu,Jun Lu,Yuzhun Lin,Xiangyun Liu
DOI: https://doi.org/10.1109/JSTARS.2022.3158905
IF: 4.715
2022-01-01
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Abstract:Characterized by complicated backgrounds, various types, large size variations, and arbitrary orientations, the detection and recognition of arbitrary-oriented objects in remote sensing images are challenging. To address the aforementioned problem, an anchor-free arbitrary-oriented object detector using box boundary-aware vectors is proposed. With the idea of CenterNet to detect objects as points, oriented object detection is achieved by predicting the center, the box boundary-aware vectors, the size, and the type of the bounding box. In the feature extraction stage of the designed architecture, Res2Net, a multiscale convolutional neural network, is used to extract feature maps of different scales and adaptively spatial feature fusion is adopted to improve the detector's adaptability to objects of different sizes. In the detector, a context enhancement module with a multibranch network is designed to enhance the contextual information of the objects and improve the detector's robustness to the complicated backgrounds. Experiments are carried on three challenging benchmarks (i.e., HRSC2016, UCAS-AOD, and DOTA) and our method achieves state-of-the-art performance with 90.30%, 89.70%, and 77.18% mAP, respectively.