Objects detection for remote sensing images based on polar coordinates

Oriented and horizontal bounding box are two typical output forms in the field of remote sensing object detection. Most present state-of-the-art detectors perform object detection in Cartesian coordinates with anchor mechanism. In these methods, the design of oriented detectors is much more complicated than the horizontal ones, because the former usually needs to devise more complex rotated anchors. In this paper, we perform object detection in polar coordinates rather than in Cartesian coordinates, and propose a novel anchor-free detector for remote sensing images. Our model, which is named as Polar Remote Sensing Object Detector (P-RSDet), takes the center point of each object as the pole point and the horizontal positive direction as the polar axis to establish the polar coordinate system. In this coordinates system, P-RSDet unifies the representation of horizontal and oriented bounding box, and makes the acquisition of oriented output form be as simple as the horizontal one. Compared with previous detectors, the design of our network is simpler and fewer output parameters need to be regressed. Experiments on DOTA, UCAS-AOD and NWPU VHR-10 datasets show that our P-RSDet achieves competitive performances on both horizontal and oriented detection fields