Edge-Feature-Based Aircraft Cover Recognition and Pose Estimation for AR-aided Inner Components Inspection

Assembly inspection before the test flight is an important measure to ensure the quality and safety of an aircraft. To improve the efficiency of the aircraft cover inspection, an algorithm based on the edge contour features for online recognition, tracking and pose estimation is proposed. We first detect the edge contour of the aircraft cover in the RGB image, and then extract the corner features as well as its location. With the prior generated XML file containing topological graph and shape descriptors of different covers as input, by adapting tracking algorithm based on DSST tracker, smoother and more accurate recognition results can be obtained. Finally, we use a set of contour corner points to estimate the camera pose, which is used to provide correct augmented reality (AR) guidance contents for inspectors. To validate our method, an AR guided system is made and tested. The experimental results show that the recognition process is very efficient, and the recognition accuracy of typical covers is 100%. The system can correctly recognize different covers and push inspection guidance information.