中文摘要：

给出一种用三个形状分布图来描述三维模型的检索算法。在模型表面任意取点，并记录每个点的法向量。将任意两点连线形成线段，计算该线段的欧几里得距离，再分别计算两端点的法向与该线段的夹角。根据两个夹角的值，将所得线段分成三个集合，并分别构造三个集合的形状分布曲线，通过对模型间三条形状分布曲线的比较，得出两个模型的相似程度，从而实现模型的相似性检索。实验表明，该算法能够较好地实现三维模型检索，检索结果比传统的形状分布算法有较大的改进。

英文摘要：

A new 3D model retrieval approach based on shape distributions was proposed, which represented a 3D shape as three D2 shape histograms. Firstly, a sufficiently large number of random sample points on surface of 3D model was taken and the normal direction of each point was recorded. Secondly, the D2 shape function was adopted to measure the distance between two random points on the surface of the model, at the same times, calculate the angles between the normals of the points and the line segment which was formed by the two random points. Lastly, according to the angles, the line segments were sorted into three sets. For each set, shape distribution histogram was calculated associating with the D2 shape distribution function. By comparing each pair of shape distribution histograms using well-known curve matching techniques and adding the three weighted results, the sum was regarded as the similar coefficient of the two models. Experiments showed that this algorithm could effectively illustrate the similar degree of the two models and as well reflecting human perceptual similarity.