中文摘要：

提取关于坐标旋转不变的特征描述符以及提高特征描述符对噪声的鲁棒性是基于内容三维模型检索技术中有待进一步研究解决的问题之一。将三维模型表面离散成三维点数据,然后使用球面调和方法对得到的点数据进行调和分析得到三维模型旋转不变的特征向量,并且能够增加算法对模型表面噪声的鲁棒性提高检索准确度。实验证明该算法在检索准确度和对噪声的鲁棒性上都有所提高。

英文摘要：

How to extract the rotation invariant feature and improve the stability of the noise is still an unsolved problem of the area of content-based 3D models retrieval.The point cloud data are extracted from the 3D models surface by casting radiative lines from the center of model.Using spherical harmonic to the point cloud data can receive a set of rotation invariant feature vector and reduce the effect of the surface noise influence and improve the retrieval precision.Experimental result shows that the proposed method receives a better retrieval performance.