摘要
车牌定位是将车牌区域从车辆图像中分割出来,是车牌识别的基础。本算法首先通过对图像进行离散余弦变换(DCT),得到DCT系数,之后计算提取出一种基于加权频率的新特征,采用自适应阈值法,实现车牌区、非车牌区的快速分割,再经平滑滤波之后,利用投影法实现车牌区域的定位。实验结果表明该算法准确率较高,有利于实现复杂背景中车牌区的快速定位。
The VLP location, which is to location the VLP from a vehicle image,is the base of the VLPR. First, the image was transferred to frequency space by DCT( discrete cosine transformation) to get the DCT coefficient. Then the plate areas can be segmented with an adaptive threshold using the weighted frequency feature extracted from DCT - compressed domain, and then after smoothing filtering plate area is located using projection algorithm. Experiments show that the algorithm can locate precisely and increases its efficiency in complexity background.
出处
《微处理机》
2010年第4期75-77,82,共4页
Microprocessors
关键词
牌照
DCT系数
加权频率
定位
Vehicle license plate
DCT coefficient
Weighted frequency
Location , the final the plate