【摘要】 介绍了神经网络分类器的基本原理,针对3类交通标志,即禁止标志、警告标志和指示标志,提出了应用神经网络分类器进行交通标志自动识别的基本方法。神经网络分类器由两层网络联结而成,前层网络由单个BP网络完成交通标志的粗分工作,后层由3个BP网络将粗分结果分别进行细分,完成识别任务。此设计结构与传统的单层分类器相比,在训练速度和识别正确率方面都有较大的提高;显然,这与神经网络在解决小规模问题时正确率高、训练速度快相符合。同时,增加新的训练样本时,只要对相应网络进行训练即可,而不必对整个网络进行重新训练。实验结果表明,基于神经网络的交通标志自动识别方法,具有很好的识别效果。更多还原

【Abstract】 The basic theory of a neural network classifier was introduced. For three kinds of traffic signs,such as indicative signs,warning signs and prohibitive signs, a method based neural network classifier was developed to recognize traffic signs. The classifier consists of two network layers. The front network is made up of a BP networks and will do a rough classification task. The back network consists of three BP networks and will fine the classification done before. The whole process completes the recognition task. Compared with a conventional single layer classifier,the training speed and the ratio of recognition of this classifier are greatly increased. It accords with the fact that the neural network has higher correctness and fast training speed when it solves small-scale problems. At the same time,when additional training samples are added, only the corresponding network needs traineing again but the whole network. The experimental results show that the automatic recognition method based on neural network has good effect.更多还原