基于样例学习的三维模型最优视图选择算法.docVIP

基于样例学习的三维模型最优视图选择算法 摘要：针对已有最优视图度量难以适用于不同类型的三维模型，所提取的最优视图往往也难以符合用户的兴趣点，本论文提出以用户定义的最优视图样例作为训练样本，以AdaBoost算法进行监督学习，通过计算输入三维模型的形状特征相似性进行分类学习，以确定该模型在最优视图样例库中对应的最优视图样例。然后，将输入模型从不同视点得到的渲染视图和该样例视图进行形状相似性对比分析，相似度最高的视图为最优视图。通过大量三维模型进行实验分析表明，本文算法得到的最优视图可以有效地逼近用户选择结果。 关键词：三维模型；最优视图选择；样例学习；相似性度量 Case-Based Learning Algorithm for Best View Selection of 3D Model Abstract: The existing metrics method for best view can not be applied widely in different kinds of 3D models. Meanwhile the best view extracted is often difficult to meet the user’s interest. In this paper, a best view case defined by user is taken as training case and the AdaBoost algorithm is used for supervised learning. The similarity of shape feature of the input 3D model is calculated and classified in order to select corresponding best view case in the database of best views for the input 3D model. Then the rendering views coming from different viewpoints for the 3D object are compared and analyzed with the best view case in shape similarity matching. The rendering view with highest similarity will be regarded as the best view of the 3D object. The experiments show that the best view selected is very consistent with the user’s subjective choices. Key words: 3D model；best view selection；case-based learning；similarity measurement 1 引言 人类视觉系统由于成像原理的局限性，仅能将三维物体映射成二维图像进行认知。但是人眼能够合理地选择视点位置，形成最优视图，实现基于单一图像的三维模型分析和理解。认知心理学方面的研究学者Verfaillie 等人更是提出了“three-quarter”这一著名概念，论证了物体的认知对视点位置具有依赖性。但由于三维模型方向的不确定性和二义性，无法确切地给出最优视图的数学定义。因此，如何模拟人类感知行为，提取最优视图成为科学界的热点研究问题。 近年来，研究人员试图通过可见面比例大小、投影面积大小、曲率、投影轮廓等几何信息，去量化人类观察物体的视觉习惯，形成一系列用于最优视图提取的评价标准[1] ADDIN EN.CITE EndNoteCiteAuthorPolonsky/AuthorYear2021 /YearRecNum182/RecNumrecordrec-number182/rec-numberref-type name=Journal Article17/ref-typecontributorsauthorsauthorO. Polonsky/authorauthorG. Patane/authorauthorS. Biasotti/author/authors/contributorstitlestitleWhatapos;s in an Image: Towards the Computation of the Best View of an Object/titlesecondary-titleThe Visual Computer/secondar