摘要
针对车牌字符图像受客观条件影响较大、样本数量不是很大、分类识别相对容易等实际情况,提出了一种对经过预处理的车牌字符图像采用SVM算法进行识别的方法。该方法将图像数据转变为文本数据作为输入样本,方法简单快速,克服了提取图像特征过程中过分依赖字符结构和图像质量的不足。并通过使用参数归整、交叉校验、选择适当的核函数等方法寻求最优参数,避免欠学习、过学习问题的产生。通过使用高速公路收费口的实拍汽车图像进行实验,验证了算法的有效性。
We present a simple method based on support vector machine （SVM） for license plate recognition. By firstly pre - processing the input images containing license plates, we obtain a set of normalized subimages, each of which contains a number, an English letter or a Chinese character. Then we simply transform these subimages into text format to avoid excessive dependency on feature extraction during recognition. Meanwhile, we eliminate outliers by scaling and perform cross - validation to find the best parameters for the SVM model. We used real color images captured at a motorway toll and LIBSVM for our experiments. Compared with the neural network method, the SVM - based method produced a higher recognition rate based on the same data set. Our experimental results also show the superiority of the SVM - based method in the case that only a small number of samples is available.
出处
《信息技术与信息化》
2007年第6期103-106,共4页
Information Technology and Informatization
