Directional Object Detection in Aerial Images

Abstract

Object detection in remote sensing consists of recognizing and localizing objects of interest in a quasi-standard bird-eye view with an arbitrary orientation on the ground plane. This naturally leads to the question: can we detect the direction of an object along with its location and object category? Herein, we propose a novel method, the directed object detector (DOD), which is capable of detecting an object’s direction together with its minimal enclosing oriented bounding box (OBB), i.e., a full object pose within the ground plane. This is integrated into an object detector to create a single end-to-end neural network. Experimental validation confirms the state-of-the-art performance on both close- and far-range remote sensing images in man-made and natural environments. Furthermore, we prove the advantage of DOD over the OBB approach in an image rectification application, which is not solvable with OBB. Code and data are available at https://github.com/molnarszilard/DOD