牌照字符识别
License plate character recognition
字符识别方法目前主要有基于模板匹配算法和基于人工神经网络算法。基于模板匹配算法首先将分割后的字符二值化,并将其尺寸大小缩放为字符数据库中模板的大小，然后与所有的模板进行匹配，最后选最佳匹配作为结果。基于人工神经元网络的算法有两种：一种是先对待识别字符进行特征提取，然后用所获得特征来训练神经网络分配器；另一种方法是直接把待处理图像输入网络，由网络自动实现特征提取直至识别出结果。
At present, the methods of character recognition are mainly based on template matching algorithm and artificial neural network algorithm. Based on the template matching algorithm, firstly, the segmented character is binarized, and its size is scaled to the size of the template in the character database, then it is matched with all the templates, and finally the best matching is selected as the result. There are two kinds of algorithms based on artificial neural network: one is to extract the features of the characters to be recognized first, and then train the neural network allocator with the acquired features; the other is to input the image to be processed into the network directly, and then the network automatically realizes the feature extraction until the result is recognized.
实际应用中，牌照识别系统的识别率与牌照质量和拍摄质量密切相关。牌照质量会受到各种因素的影响，如生锈、污损、油漆剥落、字体褪色、牌照被遮挡、牌照倾斜、高亮反光、多牌照、假牌照等等；实际拍摄过程也会受到环境亮度、拍摄亮度、车辆速度等等因素的影响。这些影响因素不同程度上降低了牌照识别的识别率，也正是牌照识别系统的困难和挑战所在。为了提高识别率，除了不断的完善识别算法，还应该想办法克服各种光照条件，使采集到的图像最利于识别。
In practical application, the recognition rate of license plate recognition system is closely related to the license plate quality and shooting quality. The quality of license plate will be affected by various factors, such as rust, dirt, paint peeling, font fading, license plate being blocked, license plate tilting, highlight and reflection, multi license plate, fake license plate, etc.; the actual shooting process will also be affected by environmental brightness, shooting brightness, vehicle speed and other factors. These factors reduce the recognition rate of license plate recognition in different degrees, which is the difficulty and challenge of license plate recognition system. In order to improve the recognition rate, in addition to constantly improving the recognition algorithm, we should also find ways to overcome various lighting conditions, so that the collected image is most conducive to recognition.