3d point cloud comparison via vertex weighted graph kernel

Creating three-dimensional models of objects in the real world has an important place in the computer vision field. However, the widespread use of three-dimensional modeling does not increase at an advanced level because it involves a sensitive, time consuming, costly and laborious process. In this study, a multi-scale vertex-weighted graph kernel is used for the fast and efficient detection and classification of three-dimensional point cloud data of real objects. Point clouds were formed for the 27 objects in this study. For point clouds, k-NN graphs are formed and the multi-scale vertex-weighted graph kernel is defined by using the eigenvalues of the Laplacian matrix. A fast and detailed classification is made about point clouds with the graph kernel function we have defined.