中文摘要：

针对二次误差测度算法存在尖端特征消失、局部过度简化等缺陷，提出了基于特征矩阵的二次误差测度算法用于网格简化．通过将顶点曲率和边长引进该特征矩阵以优化误差度量，模型中各顶点便易于区分，于是具有明显几何特征区域的顶点误差度量能够被提高．这样，边折叠的顺序可以方便的得到调整，使得模型中的突出特征更多的被保留下来．仿真结果表明，本算法在保持了二次误差测度算法计算时间短、运行效率高的同时，也克服了网格分布过于均匀、无法突出模型重要特征的缺点．

英文摘要：

Contraposing the some deficiencies from the algorithm based on quadric error metrics （QEM） , such as neglect of some cusp features and excessive simplification in some parts of the model, a QEM based on eigenmatrix was proposed for mesh simplification. Through introducing the curvature and edge length of a ver- tex into this new QEM to optimize error metrics, the eigenmatrix can easily distinguish the vertexes of a model, improve the error metrics of the vertexes in the areas with obvious geometric features. Therefore, the folding sequences of edge can be adjusted, so that some sharp features of the object can be preserved. Comparing with QEM algorithm, the simulation results show that the proposed approach can not only keep merits such as high executing speed and running efficiency, but also overcome the shortcoming to easily lose some important features of models because of mesh distribution too equality.