关键词: 超二次曲面建模, 深度图像分割, 部件识别, 演化计算

Abstract: In range image processing，the research on modeling and recognition for scattered data points is more universality，and it has became the focus of attention in computer vision.For the characteristic of discrete and irregular 3D data points，the existent problems of superquadric parametric fitting，multi-object scene segmenting and parts recognizing are analyzed through investigating the theory and method of superquadric modeling，segmentation and recognition，including the application of evolutionary computation in 3D modeling and recognition.It is proposed to enhance the capability of superquadric representation and to make use of the basic elements of superquadrics for scene modeling and segmenting in future work，and it also is put forward to bring the theory of swarm parallel evolutionary and relational match into the superquadric modeling and recognition.The research aims to explore a scheme of 3D modeling and recognition which is effective and practicable.

Key words: superquadric modeling, range image segmentation, parts recognition, evolutionary computation