摘要
提出了一种基于支持向量机(SVM)的改进车牌识别方法。对细化处理后的字符采用网格、水平投影与垂直投影密度的特征提取方法,保证了字符整体与局部特征,同时也使特征向量集的维数充分低。结合3种特征提取方法得到的特征向量集,采用SVM进行车牌号码识别。对于易混淆字符,提出了根据各自的特征进行2次识别的算法,该算法有效解决了易混淆字符误识别的问题。实验结果表明,该算法鲁棒性好、抗干扰能力强、识别率达到了98.58%。
An improved recognition method of license plate based on support vector machine （SVM） is pro- posed in this paper. The features of grid, horizontal projection and vertical projection density are extracted after char- acter thinning treatment. The feature vector set of three features has overall and local features of the characters, and low dimension. Based on SVM, plate numbers are recognized through feature vector sets. For confusing characters, a second recognition algorithm is put forward according to their respective features, which solves the problem of confus- ed character recognition efficiently. Experiment results show that the algorithm has good robustness, strong anti-jam- ming ability, and a high recognition rate to 98.58%.
出处
《电子科技》
2013年第11期22-25,共4页
Electronic Science and Technology
关键词
车牌识别
SVM
特征向量
网格
投影
细化
license plate recognition
SVM
feature vector
grid
projection
thin
