A hybrid method for skeleton extraction on Kinect sensor data: Combination of L1-Median and Laplacian shrinking algorithms

The distinction of three-dimensional objects is one of the main challenges in computer graphics and computer vision. Distinguishing and recognizing between objects and shapes which are frequently encountered in everyday life is an important problem. In this work, a robust curve skeleton extraction algorithm is introduced on point clouds data for 3D real objects. The curve skeleton of the 3D object is a discrete geometric and topological representation of 3D shapes and maps spatial relationship of the geometric parts according to the graphical structure. Skeleton structure is the integrated stage of an average point clouds data obtained from the existing point cloud. The presented algorithm works on the average metric values of the point clouds and compensates for some missing point clouds that can be found in point clouds generated from objects. The developed method uses a combination of L1-Median and Laplacian shrinking algorithms. Moreover, a curve skeleton can be extracted on the partially deformed point cloud. Thus, curve skeleton becomes convenient to define and process objects used in the geometric modeling. The resulting skeletal structure provides a method of object recognition that can cope with objects having complex geometry.