Multiple circle recognition and pose estimation for aerospace application

A novel recognition of ellipses and relative pose estimation algorithm is presented based on surface of revolution (SOR) in this paper for images of multiple ellipses. Firstly, the contours following method is performed on image to detect ellipses. Secondly, candidate ellipses projected by circular cross section of a SOR are obtained based on parallel constraint and vertical constraint in the image. Finally, relative pose between target and camera is calculated by these candidate ellipses. Experimental results indicate that the method perform well in recognition and estimation. The precision of recognition is higher than 97% for synthetic images corrupted by 0 ∼ 16% salt-and-pepper noise. The absolute error of pose angle is 1°,and the absolute error in axis z and other axis are less than 80 mm and 15 mm, respectively, when measure distance is less than 10 m.