基于BP网络分类器的交通标志识别

Traffic Sign Recognition Based on BP Network Classifier

针对中国全部3大类116个交通标志,即禁令标志、指示标志、警告标志,用BP网络实现分类功能.实验中使用了3种测试集,即加高斯噪声、水平扭曲和日本交通标志实景图,对BP网络的分类性能进行了测试.实验结果表明:用BP网络实现交通标志粗分类功能的效果是比较理想的,对交通标志的颜色失真和形状失真均具有较好的容错性和鲁棒性.

In this paper, three types of traffic signs （i.e. warning sign, interdictory sign, and indicatory sign, totally 116 signs） are further classified by the classification function of BP network. In our experiment three testing sets, namely, Gauss Noise signs, level twisted signs and Japanese real signs are employed to test the performance of BP network classification. The experimental result shows that BP network is highly fault-tolerant and robust for dealing with the sign＇s distortion of color and shape, and thus it is ideal for application of traffic signs classification.