模型驱动下的形状分布算法

Model-driven Shape Distributions

三维模型的特征提取方法中,形状分布算法简单,具有强的不变性和鲁棒性,但其抽样点不具有代表性,影响了系统的检索精度.为增强抽样点的有效性,我们改进抽样策略,首先引入模型复杂度描述参数,扩大抽样点集;然后利用可分离离散小波变换,逐层分解样点集,增强其有效性,并使最终样点数相同;最后使用形状分布算法进行三维模型的特征提取,并应用在三维模型检索上,对比实验结果表明,改进后的算法提高了系统的检索精度.

Shape distribution is a simple method in feature extraction for three-dimensional model which has a strong invariance and robustness. But for its sample＇s snon-representative, the retrieval precision decreases. To enhance the effectiveness of samples, we improve the sampling strategy： first, determine the number of pre-sampling point according to the model＇s file level; then use separable discrete wavelet transform to enhance the effectiveness of samples set and compress the data set; fmaUy shape distribution algorithm is carried out. The simulation results illustrate the modified algorithm＇s rightness.