论文部分内容阅读
针对较低分辨率的车牌汉字识别问题提出了一种改进的模板匹配算法。在扩展的二值图中寻找字体边缘,并截取灰度原图进行匹配;采用组合模板减少目标倾斜、伸缩的影响;根据大样本统计自动选择模板和匹配阈值。经交通路口实拍图像采样测试,整体识别率高于97%,在P41.8GCPU、512MB内存的台式计算机Matlab环境下识别一个汉字平均耗时小于0.06s。本文算法对低分辨率车牌汉字的识别率高,处理速度满足实时性要求,具有实际应用价值。
An improved template matching algorithm is proposed for the problem of lower resolution Chinese character recognition. Find the edge of the font in the extended binary image, and cut off the gray-level original image to make a match. Use the combined template to reduce the influence of the target tilt and telescopic. Select the template and matching threshold automatically according to the large sample statistics. The image taken by the traffic jerk real shot test, the overall recognition rate of more than 97%, P41.8GCPU, 512MB of memory in the desktop computer Matlab environment to identify a Chinese characters on average less than 0.06s. The algorithm of this paper has high recognition rate of low-resolution license plate Chinese characters, processing speed to meet the real-time requirements, and has practical application value.