基于多分类器--并行计算的车牌识别算法

针对已有车牌识别中技术存在的不足，提出了一种多分类器——模板匹配和神经网络并行计算的识别算法，这一方法对于汉字、英文和数字混杂、数字的识别，分别采用粗分类和面向汉字的双进程计算方法、面向字母的双进程计算方法、简单的数字神经网络方法。这些方法的采用可以缩小检索范围，充分利用模板匹配和神经网络算法各自的识别优点，提高车牌字符识别准确率，并进一步提高运算速度。

Recognition Algorithm of License Plate Based on Multi-classifier and Parallel Computing

The traditional character recognition technology has some deficiencies in the recognition of vehicle license plate. This paper introduces a new recognition algorithm with multiclassifier-template matching and parallel computing of neural networks. For the Chinese characters recognition, the coarse classification and double course computing method facing Chinese characters are used. For the recognition of mixing of English and numbers, the coarse classification and double course computing method facing letters are applied. For the number recognition, simple digital BP neural network is adopted. The methods can reduce the searching range, take full advantage of recognition merit of template matching and BP neural network algorithm, improve accuracy rate of recognition for license plate characters, and increase operation speed further.