Abstract: In order to solve the problem of recognition difficulty for the absence of uniform standard in the rail power supply contact detection, a real-time pole number recognition algorithm based on adaptive shape context feature is proposed. The pole number is recognized and located by Shape Context matching algorithm with the known template characters and pole image, and then the correct pole number is obtained by the adaptive strategy algorithm. The experimental results show that the number of poles with different shapes and sizes can be identified by less configuration with good engineering practice significance, and the accuracy is more than 95%.

Key words: shape context, pole number recognition, adaptive

摘要： 为了解决铁路供电接触网检测中杆号没有统一标准而造成识别困难的问题，提出了一种基于自适应形状上下文特征（Shape Context）的实时杆号识别算法。该算法利用已知的模板字符与杆号图像通过Shape Context匹配算法进行杆号识别和定位，然后通过自适应策略算法得到正确的杆号。实验结果表明，对于不同形状，不同大小的杆号牌，辅以较少的配置就可以实现杆号识别，准确率在95%以上，具有良好的工程实践意义。

关键词: 形状上下文特征, 杆号识别, 自适应