一种基于HSV颜色空间和SIFT特征的车牌提取算法

摘要

为了克服SIFT算法直接应用在车牌提取中表现出来的执行时间过长、误配率高的缺陷,提出了一种基于HSV颜色空间与SIFT特征的两级车牌提取算法,先使用HSV颜色空间确定车牌的候选区域,进行快速粗定位,再使用SIFT算法对候选区域进行精确定位与倾斜校正,在精确定位的同时也完成了对车牌汉字的辨识.这种方法不仅减少了SIFT特征的计算量,而且也避免了复杂背景对于SIFT特征匹配的干扰,大大提高了匹配准确率.最后通过编程实验证实本算法有良好的性能.%In order to overcome the drawbacks such as slow execution speed and high mismatch rate in directly apply SIFT feature to vehicle plate extraction,this paper proposed a two steps vehicle plate extraction method based on SIFT feature and HSV color space. First roughed location the candidate region use HSV color space in a short time, then adopted SIFT feature to locate the plate precisely and correct the tilt,in the meantime accomplished the recognition task of the Chinese character on the vehicle plate. Through the approach not only reduced the computation of SIFT feature, but also avoided the background interference, so can match the features fast and correct. Experimental results also show that the method has a good performance.