摘要
针对三维模型的分类问题,提出了一种基于统计特征量和Markov模型的分类算法。该算法对预处理后的三维模型进行几何切分,并提取切分后每块的统计特征。对三维模型各分块进行一定顺序的观测,可以获得由各分块的统计特征量构成的Markov模型的伪时间序列。再对不同类模型进行训练并得到各类模型对应的Markov模型参数。最后定义模型间的相似度度量,获得三维模型的分类结果。实验表明该算法在绝大多数类别的模型上分类效果较好,准确率达到90%以上。
This paper proposes a new method of 3D object classification based on statistical feature and Markov model.It spatially divides the 3D model of objects into different parts and then extracts statistical feature from each part.The different parts in spatial domain are observed as a certain sequence,pseudo time series of Markov model composed of different parts statistical features are acquired.The model parameters are learned for different class from the training data.In recognition,the output class label of a 3D model is the one that maximizes the joint likelihood.The experiment achieves better results in most cases with accuracy of 90%.
出处
《计算机工程与应用》
CSCD
北大核心
2011年第4期157-159,共3页
Computer Engineering and Applications
基金
国家自然科学基金(No.50846021)
校内研究生创新教育项目(No.09YJC32)
校内科学研究基金(No.ND0936)~~
