摘要
现有车牌定位算法在车牌颜色和车辆颜色接近时,算法的有效性会下降。鉴于这种情况,提出一种基于边缘信息的车牌定位方法,主要根据车牌区域边缘信息丰富的特点,结合脉冲耦合神经网络,与Sobel算子对输入彩色图像实行多通道边缘检测与分割,再结合投影信息进行定位。实验结果表明,该方法对不同光照度和不同颜色组合的车牌图像的定位具有良好的鲁棒性。
The conventional methods for License recognition system often suffer declining accuracy when license plates are contaminated or blurred.In order to make an improvement,a methods based on edge statistics are presented,which are stem from the fact that there are constantly large gradient values in license plate areas.Experiment result indicates that this method assumes a robust performance on images with license plate of diverse color and images with various illuminations.
关键词
边缘投影
脉冲耦合神经网络
车牌定位
edge projection
Pulse Coupled Neural Network
license localization
