智能装配系统中神经视觉的三维物体识别

根据智能装配系统的实际要求，提出了一种利用神经视觉进行三维物体识别的理论和方法，在利用立体象对重建物体的三维外形的基础上，建立物体的区域图，利用物体的三维矩及其不变性来构造代表物体的特征矢量．采用ＡＲＴ２神经网络构成神经网络分类器，把物体的特征矢量作为神经网络分类器的输入，从而对物体进行识别或分类．这种识别或分类方法可以在线学习，能满足智能装配环境下连续作业的要求

3D Object Recognition Using Neuro Vision in Intelligent Assembly System

A 3D object recognition method using neuro vision is presented based on the research of intelligent assembly system. In this method, the stereo image pairs are obtained by using vision system first. The 3D information is extracted from the stereo image pairs, and then the 3D configuration and surfaces of object are reconstructed, so that the range map of the object can be obtained. The feature vectors representing objects to be recognized are constructed by using 3D moment and its invariant. The ART2 neural network is adopted to construct classifier. After the input of the feature vectors into the classifier, the objects can be recognized or classified. This classifier can learn knowledge on line, so that it can be used in continuous operation in intelligent assembly system. It is proved through experiment and actual applications that the method presented in this paper are correct. It can be used in intelligent assembly system and in other fields.