复杂环境下多特征融合的车牌定位分割算法

Multi-Feature Fusion Algorithm For License Plate Loaction Segmentation In Complex Environment

提出一种基于几何、颜色、边缘等多特征融合的车牌定位分割算法.首先,在车牌定位过程中,利用特征点检测和形态学处理结合的方法,综合车牌的几何、字符纹理和颜色等特征进行粗定位;其次,利用灰度投影技术,对车牌字符进行精确定位;最后,利用灰度投影自适应分割手段进行车牌字符分割.实验表明:该算法的定位分割成功率达到95%以上,明显高于其他对比算法,能较准确地定位并分割字符.

A license plate location and segmentation algorithm based on multi-feature fusion such as geometry,color and edge was proposed.First of all,in the process of license plate positioning,the method combining feature point detection and morphological processing was used to comprehensively locate the license plate based on its geometry,character texture,color and other characteristics.Secondly,the grayscale projection technology was used to accurately locate the license plate characters.Finally,the gray projection adaptive segmentation method was used for segmentation.The experiment showed that the success rate of localization segmentation was more than 95%,which was obviously higher than other comparison algorithms,and could locate and segment characters more accurately.