摘要
提出一种直接使用３Ｄ信息的目标识别方法。３Ｄ不变性矩可作为任意３Ｄ形状的表示特征且具有平移、旋转和尺度变化的不变性。人工神经网络的训练过程对应于主动式机器视觉的内部建模过程。
In this paper we propose a method for object recognition from a range image. The 3-D moment invariants are invariant against translation, rotation and scaling and are suitable for the representation of any shape. The training process of the BP artificial neural networks corresponds to the modeling in the active computer vision. This method is suitable for 3-D object recognition in active computer vision.
基金
863项目资助
关键词
不变性矩
目标识别
神经网络
水下机器人
D information
moment invariants
active computer vision
object recognition
artificial neural networks
