Accurate Monocular Pose Measurement for Aircraft Transportation Monitoring

Comprehensive statue monitoring is of great significance for aircraft safety and efficient transportation. This paper proposes an accurate and efficient monocular pose measurement method for aircraft transportation monitoring. The proposed method firstly models the aircraft using a sparse keypoint set for simplicity. Then, based on the simple representation, a trained neural network model is adapted to detect the keypoints from the input monocular image to build 2D-3D correspondences for pose measurement. Coarse pose estimation is got by solving a Perspective-n-Point problem based on the 2D-3D correspondences. For accurate pose measurement, this paper uses the contour feature to refine the pose measurement. The refinement is achieved via gradient orientation based template matching. Experimental results indicate that the proposed method achieves accurate and efficient monocular pose measurement for aircraft transportation monitoring.