基于二值图像层次识别字符特征的融合算法

针对图像二值化过程中,阈值选择不当容易造成字符笔画粗细方向的偏差,从而影响统计特征的问题,提出了一种基于二值图像层次识别字符的融合算法.该算法采用神经网络作为分类器,充分利用字符图像中点阵、特征线和霍夫矩等各组特征向量的互补作用,以提升识别系统的性能.实验结果表明:该融合算法的识别率可达97%以上,平均运行时间小于50 ms,能够满足车牌识别系统实时要求.

Fusion algorithm of distinguishing characteristic of character based on binary images

Directed at the negative effect on the character of the stat imposed by the deviation in the character stroke of a Chinese character due to the wrong choice of threshold in binary conversion of image, the paper proposes the fusion algorithm distinguishing characteristic of a character based on binary image arrangement to statistics characteristic. The fusion algorithm improves the distinguishing ability of the system by adopting neural networks as classification implement and utilizing the mutual effect by character image, lattice, characteristic line and Huo Fu regulation. As the experimental results show, the algorithm, with the efficiency of above 97% and the average value of running of less than 50 ms, meets the real time demand of the vehicle license plate recognition system.