一种微光下车牌图像的去噪和定位方法

A Method of Vehicle License Plate Image De-noising and Location in Low Light Level

车牌识别系统是智能交通的重要组成部分,而车牌定位是车牌识别系统的关键部分,它直接影响到后续的字符分割和识别。根据微光下车牌图像含椒盐噪声的特点和小波变换的集中特性,对微光车牌图像进行小波变换处理,用软阈值法对小波系数进行阈值处理,将得到的高频小波系数进行重构,得到去噪图像;最后对去噪后的车牌图像利用边缘检测法和投影法进行车牌区域定位。实验结果表明:每张车牌平均定位时间为1.78 s,定位成功率达98%。

License plate recognition system （LPRS） is the most important part of the intelligent transportation system （ITS）. The license plate location is the key part of the LPRS. It affects the latter character segmentation and recognition directly. According to the characteristic of the license plate image which contained the slat and pepper noise in low light level （LLL） and the concentration characteristic of wavelet transform, wavelet transform was used to process vehicle license plate image in LLL. The soft-threshold meth- od was used to process the wavelet coefficients to reconstruct the high frequency coefficients, and the de-noising image was obtained. The methods of edge detection and projection were used to locate the vehicle license plate area for de-noising vehicle license plate image. The experimental results show that the average location time of each license plate is 1.78s and the rate of accuracy is 98%.