摘要
在恶劣的雨天环境中,随机分布了大量快速运动的雨滴,造成目标物体与背景光线的反射和折射,使得图像对比度降低、成像模糊、细节信息丢失,从而降低成像系统获取的汽车图像质量,影响了车牌检测的效果。针对该问题,提出了一种基于相对总变差模型和频域处理的车牌检测方法。首先,基于相对总变差模型分解图像,可以得到包含雨线的纹理图;然后,将纹理图作离散傅里叶变换后,在频域内有效地对雨线进行分析和滤除,对去除雨线后的纹理图与结构图重构得到去雨后的汽车图像;最后,采用基于局部统计滤波的方法对去除雨线后的图像进行车牌检测。试验结果表明,该方法可以有效地检测出雨天条件下的车牌,并且车牌检测准确率高、耗时短,具有实际工程应用意义。
In the harsh environment of rain,a lot of fast moving and randomly distributed raindrops cause the reflection and refraction of the light on the objects and background,so the image contrast is reduced,the image becomesblurred,and the detail information may be lost.Thus,the quality of images of vehicle obtained by the imaging system is worse,and the license plate inspection is greatly influenced.To deal with this problem,the license plate inspection method based on total variation and frequency domain processing is proposed.Firstly,the image is decomposed based on the model of relative total variation for extractingtexture image with rain lines.Secondly,the texture image is transformed by discrete Fourier transformation,so the rain lines are effectively analyzed and filtered out in frequency domain; then,the vehicle image is obtained by reconstructing the structural image and texture image without raindrops.Finally,the license plate inspection is conducted by adopting the method based on local statistic filtering and the image with rain removing.The experimental results show that this method can detect the license plate effectively under rainy conditions,and it is of higher accuracy and shorter time consuming,and providing practical applicable significance in engineering.
出处
《自动化仪表》
CAS
2017年第4期76-80,共5页
Process Automation Instrumentation
关键词
智能交通
车牌检测
频域处理
傅里叶变换
滤波
图像
Intelligent transportation
License plate inspection
Frequency domain processing
Fourier transform
Filtering
Image
