基于2D-LDA的车牌字符识别  

License Plate Character Recognition Based on 2D-LDA

作　　者：周洪毅[1] 

机构地区：[1]南京理工大学自动化学院,江苏南京210094

出　　处：《数字技术与应用》2016年第6期97-99,共3页Digital Technology & Application

摘　　要：传统的基于LDA的字符识别需要将图像向量化,这会造成协方差矩阵维数过大和奇异问题,而基于2D-LDA的识别算法能够克服上述传统算法的缺陷。首先介绍了2D-LDA算法的原理;然后,在车牌字符数据集上测试了算法的识别率;最后,与多层感知机神经网络做了对比,表明2D-LDA算法有较高的识别率。The traditional character recognition based on LDA requires the image vectorization processing, which will bring about a too large covariance matrix dimension and singular problems, but 2D-LDA based recognition algorithm can overcome the shortcomings of the traditional algorithm described above. First, this paper introduced the principle of 2D-LDA algorithm. Then it tested the recognition rate of the algorithm on the license plate character data set. At last, it gave a comparation between 2D-LDA algorithm and MLP neural networks. The comparation indicated 2D-LDA algorithm has a higher recognition rate.

关 键 词：字符识别 2D-PCA 2D-LDA 多层感知机神经网络 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]