Three-dimensional Model Segmentation Based on Improved Random Walk Segmentation Algorithm

In1 order to solve the problem of the poor precision and instability of the existing 3D model segmentation algorithm, an improved random walk algorithm is proposed for 3D model segmentation. First, an empowerment model is constructed for the input 3D model, and the transformation matrix is obtained. Then, the model is marked as seeds by interaction and divided into k meaningful regions, obtaining the k initial distributions. Finally, the final probability distribution is obtained by iterating to convergence, and the maximum posterior probability is used to the model punctuation, so as to realize the 3D models segmentation. Experiments show that the method has achieved good segmentation results on irregular face models and the Terracotta Army fragment model. Compared with other methods, the algorithm has better performance, fewer iterations and shorter running time.