基于模板匹配和神经网络车牌识别技术的研究

Research on license plate character recognition based on template matching and neural network

提出了一种基于模板匹配和神经网络相结合的车牌字符识别算法.采用二维经验模式分解算法(BEMD)对图像进行去噪处理,用Sobel算子进行边缘检测,使用累计直方图和低分辨率图进行特征提取,利用模板匹配法对车牌进行粗识别,对于模板匹配不可识别或难于识别的字符改用BP神经网络进一步识别.实验结果表明,车牌的识别率和识别速度都有所提高.

In this paper, an algorithm of license plate character recognition based on template matching and neural network was proposed. The bidimensional empirical mode decomposition （BEMD） was applied for image denoising. The Sobel operator was used for detecting image edge, and then image feature extractions were realized by the cumulative histogram and low resolution image. Finally, parts of the license plate characters were roughly recognized by template matching. The rest which cannot be recognized by template matching can be ob- tained using Neural Network. Experimental results showed that the algorithm could improve the recognition rate and reduce the recognition time.