论文部分内容阅读
在本文中,我们用统计模式识别的方法分析了目前在模式识别中得到广泛应用的多层BP神经网络,揭示了具有线性输出单元的多层BP神经网络用作特征提取器和分类器时具有良好性能的原因。同时,我们设计了一个使用多层BP网络作为特征提取器和分类器的、普适的工件识别系统。对不能识别的样本,采用模糊推理技术,把传统的直观特征识别结果和多层BP网络结果在特征级上融合,提高系统的性能。
In this paper, we analyze the multi-layer BP neural network, which is widely used in pattern recognition, using the statistical pattern recognition method. It is revealed that the multi-layer BP neural network with linear output unit has the following features when used as feature extractor and classifier: Reason for good performance. At the same time, we designed a universal workpiece recognition system using multi-layer BP network as feature extractor and classifier. For unrecognized samples, the fuzzy reasoning technology is adopted to fuse the traditional visual feature recognition results and the multi-layer BP network results at the feature level to improve the system performance.