A Minimalist Approach to Type-Agnostic Detection of Quadrics in Point Clouds

摘要: This paper proposes a segmentation-free, automatic and efficient procedure to detect general geometric quadric forms in point clouds, where clutter occlusions are inevitable. Our everyday world is dominated by man-made objects which designed using 3D primitives (such as planes, cones, spheres, cylinders, etc.). These also omnipresent industrial environments. gives rise the possibility of abstracting scenes through primitives, thereby positions these an integral part perception high level scene understanding. As opposed state-of-the-art, tailored algorithm treats each primitive type separately, we propose encapsulate all types single robust detection procedure. At center our approach lies closed form fit, operating both primal & dual spaces requiring low 4 oriented-points. Around this design novel, local null-space voting strategy reduce 4-point case 3. Voting coupled with famous RANSAC makes orders magnitude faster than its conventional counterparts. first method capable performing generic cross-type multi-object difficult scenes. Results on synthetic real datasets support validity method.