Aircraft Component Detection Based on 3D Object Recognition and Relative Position Estimation

An aircraft component detection method based on 3D recognition and relative position estimation is proposed in this paper. Direct detection of components that are small parts of an object is difficult for the lack of distinctive features. Since relative position of a component to the main axis of the plane is invariant to 3D transformation, major direction vector is proposed to find search region that encloses interesting parts. Major direction vector is parallel to the projection of the main axis of a plane in 2D images. 3D recognition based on shape features is applied to estimate pose of a plane. Fourier Descriptors are applied to extract features. The detection in an image is reduced to the search region after the two steps. A detection rate of 84% is achieved in the search of landing gear.