基于神经网络的电表数字识别技术研究

The research on identification technology of electric meter based on neural network

提出了基于BP神经网络的机械式电表数字自动识别方法,首先通过预处理自动定位电表图像中的数字区域并实现单个数字的切分,然后对每个数字图像提取一组具有较高区分度且计算简单的典型网格特征,最后设计BP神经网络作为数字分类器,实现电度表显示值快速自动识别,该研究获得了电表数字正确识别率98.5%的结果,表明该系统具有较强的鲁棒性。

This paper presents an automatic identification system for mechanical electric meter based on the BP neural network.The rapid automatic identification of electric meter is realized by digital classifier based on BP neural network,whose design comes from a group of typical grid characteristics with high degree of distinction and simple calculation extracted from each image of single digital segmentation through pretreatment of the digital area image of automatic positioning electric meter.The study shows that the recognition accuracy is as high as 98.5%,which indicates that the system has strong robustness.