Abstract: In the 3D object recognition system，this paper presents to combine the lower order of Hu’s moment invariants and affine moment invariants together as features of 3D objects，then these features are presented to the modified BP neural network for 3D object recognition.The theoretical and experimental analyses prove that using the combination of Hu’s moment invariants and affine moment invariants as features to classify 3D objects can achieve better recognition performance than only using Hu’s moment invariants.If the combination of Hu’s moment invariants and affine moment invariants is further processed by principal components analysis，system recognition performance can be improved greatly and network training time can be reduced.

Key words: 3-D object recognition, Hu’s moment invariants, affine moment invariants, BP neural network, principal components analysis

摘要： 在三维物体识别系统中，提出将三维物体的Hu不变矩和仿射不变矩两者的低阶矩组合作为三维物体的特征，结合改进的BP神经网络应用于三维物体的分类识别。理论分析和仿真实验表明组合这两种矩特征进行物体识别，性能优于单独使用Hu不变矩，如果进一步对这两种组合的矩特征进行主成分分析处理，可显著提高系统识别性能，并减少网络的训练时间。

关键词: 三维物体识别, Hu不变矩, 仿射不变矩, BP神经网络, 主成分分析