Three-dimensional grid model segmentation method and system based on energy optimization

Abstract

The invention discloses a three-dimensional grid model segmentation method and system based on energy optimization. The method comprises: step 1, using an energy function, concavity and convexity anddiscrimination characteristics for initially marking an input model, and constructing a segmentation area; and step 2, obtaining an accurate segmentation ring by using an improved corrosion algorithm,an energy minimization principle and curvature similarity, the improved corrosion algorithm ensuring that the connectivity of the original segmentation area and the internal area is not changed whenthe segmentation area is refined; the energy minimization principle and curvature similarity ensuring that the segmentation lines grow according to reasonable segmentation boundaries and follow the structural characteristics of the existing segmentation lines; and step 3, optimizing the segmentation ring by using a regional connectivity principle and a Dijstra single-source shortest path algorithm. A part of invalid segmentation line segments are removed through regional connectivity, and a Dijstra algorithm enables sawtooth parts on a segmentation ring to be smooth. Compared with most singlefeature segmentation algorithms, the method is higher in segmentation precision and higher in robustness, and the segmentation result is more easily accepted by people.

Description

technical field [0001] The invention belongs to the field of computer vision, and in particular relates to a three-dimensional grid model segmentation method based on energy optimization. Background technique [0002] In recent years, with the rapid development of computer vision and the improvement of computer hardware level, 3D models have been widely used in many fields of life, including film and television animation, 3D games, virtual reality, automatic driving, industrial production and other aspects. Therefore, 3D model processing has become a research hotspot in recent years. The core requirement of 3D model segmentation is to establish the connection between abstract data and the real world by processing the 3D model, so as to obtain the perception and understanding of the real world. Based on this research hotspot, the present invention conducts in-depth exploration and research on the segmentation of three-dimensional mesh models. [0003] Mesh segmentation is t...

AI Technical Summary

Problems solved by technology

This type of segmentation method is still affected by different geometric features on the segmentation results

By introducing the method of machine learning, the dividing line constructed by using various geometric features as comprehensive features is more in line with human visual perception, but this method requires a large amount of data set training in the early stage, and is not suitable for scenarios without prior conditions

Images