基于光照补偿的夜间车道线检测方法

Nighttime lane line detection based on illumination compensation

针对夜间环境下图像整体亮度较暗,光照不均匀造成车道线不易检测的问题,提出一种基于光照补偿的夜间车道线检测方法。将多尺度retinex(Multiscale Retinex,MSR)算法的光照补偿图像和经验模式分解(Empirical Mode Composition,EMD)的内蕴模式分量图像融合以增强图像的光照和对比度。提出基于几何约束的车道线筛选方法,利用基于密度的空间聚类方法对车道线进行分步聚类,消除夜间因光照产生的虚假车道线。实验结果表明,与现有方法相比,所提方法夜间车道线检测准确率有较大提高。

In order to solve the problem that the image is dark and the illumination is not uniform,a nighttime lane detection method based on illumination compensation is proposed.Illumination compensation images of Multiscale Retinex(MSR)algorithm and images of intrinsic mode components of Empirical Mode Composition(EMD)were superimposed to enhance illumination and contrast of the images.Lane lines were extracted by Angle features based on geometric constraints,and then the spatial clustering method based on density was used to cluster the lane lines step by step to eliminate the false lane lines caused by light at night.Experimental results show that the proposed method is more accurate than the existing method in lane line detection.