Superquadrics Model-based 3D Object Localization Algorithm

For the pose parameter estimation problem of 3D objects in an unorganized point cloud, a 3D object localization algorithm based on superquadrics model is proposed. A normalized radial Euclidean distance of a space point from the 3D object under arbitrary pose is defined by using the part-based superquadrics model of the 3D object. Then a nonlinear objective function for the 3D object pose estimation is established according to the mean square distance between the object surface points and the object in the point cloud, as well as the surface point number and interior point number of the object. By this means, the object localization problem is transformed into an optimization problem of the objective function. Then the invasive weed optimization (IWO) algorithm is adopted to optimize this objective function, and the obtained optimal solution is used as the estimation value of the 3D object pose. Experimental results demonstrate that the proposed algorithm can yield accurate object localization results with a good consistency of pose parameters, and effectively suppress the influence of measurement noises on measurement results.