面向车道线图像识别的多种滤波方式对比分析

Comparable analysis of multiple filtering methods for lane line image recognition

针对图像预处理的结果会影响车道线识别精度和速度的问题,文中选取二维小波包分解滤波、中值滤波和锐化滤波对某一汽车常用工况下的车道线图像分别进行滤波降噪,采用Sobel算子对未经滤波处理和经过3种滤波处理的车道线图像进行了边缘检测。结果显示二维小波包分解滤波处理后的图像能识别的车道线特征点数目最多,中值滤波处理后的图像能识别的车道线特征点占比最高,锐化滤波处理后的图像识别速度最快。

With regarding to the problem that results of the image pre-processing wilt affect the identification of lane line in its accuracy and speed,this paper chooses the decomposition filtering, median filtering and sharpening filtering of two-dimensional wavelet packet to filter the noise of the lane line image for the particular vehicle under common operating condition and the Sobel operator is used to test the the edge of the lane line images of those which has non-filtering processing and experienced three filtering processes.The results show that images,which have been processed by the decomposition filtering by the two-dimensional wavelet packet,identify the largest number of lane line feature points;images,which have been processed by the median filtering,identify the highest proportion of lane line feature points;images,which have been processed by the sharpening filtering have the highest identification speed.