Application of neural networks in segmentation of range images

A Kohenen self-organizing neural-network-based approach to range image segmentation is presented. Orthogonal Zernike moments are computed locally to extract discriminatory surface features. These features are fed to a 1-D Kohonen neural network (NN) to provide final grouping pixels. Preliminary results show that NN-based clustering outperforms or at least performs as well as traditional clustering methods of segmenting range data