摘要
对目标进行合理的描述是进行目标识别的基础.对于三维目标来说,目标表面梯度反映了目标形状的三维信息.对目标表面梯度进行描述也就是对目标的三维形状进行描述.基于这一思想,本文提出了梯度矩的概念,给出了梯度矩的定义,并讨论了梯度矩的不变性.利用梯度矩可以把三维空间中的问题转化到二维空间中去,这样不仅显著减少了计算复杂度,也可以利用成熟的二维矩理论进行研究.
Reasonable description of objects is the foundation of object recognition. Because the surface gradient reflects the 3D information of object shape, the description of object surface gradient means the description of the object shape itself. In this paper the conception and definition of gradient moment is proposed, and its invariant attribution is discussed. By using this proposed gradient moment, the problem in 3D domain can be transformed into 2D domain. Thus, not only the complexity of moment computation is reduced, but also the traditional 2D moment theory can be used in researches.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2004年第3期352-356,共5页
Pattern Recognition and Artificial Intelligence
基金
中国博士后科学基金(No.2003034156)
关键词
梯度
矩
矩不变量
目标描述
Gradient
Moment
Moment Invariant
Object Description
