全景鸟瞰图像中车道线颜色识别方法

Lane line color recognition in panoramic bird view images

车道线颜色是一种特殊的车道属性信息,是驾驶员文明驾驶和安全驾驶决策的重要依据,而现有的研究只是基于视觉的车道线检测,并未对其进行颜色识别,为此提出一种全景鸟瞰图像中车道线颜色识别方法。首先对全景鸟瞰图像中检测到的车道线进行兴趣点提取,然后将兴趣点三维RGB像素信息转换为二维信息数据,最后利用mean-shift算法聚类寻找质心,依据不同颜色的二维信息中的分布特征识别车道线颜色。实验结果表明该方法对全景鸟瞰图像中的车道线颜色识别有很好的正确率。

Lane color is a special lane attribute information,which is an important basis for driver′s civilized driving and safe driving decision.However,the existing research is only based on visual Lane detection,and does not recognize its color.This paper presents a lane line color recognition method in panoramic bird′s-eye view image.Firstly,interest points are extracted from the lane lines detected in the panoramic bird′s eye view image.Secondly,the three-dimensional RGB pixel information of interest points is transformed into two-dimensional information data.Finally,the center of mass is found by means of mean-shift clustering algorithm,and lane line color is recognized according to the distribution characteristics of two-dimensional information of different colors.The experimental results show that the method has a good accuracy for lane color recognition in panoramic bird′s-eye view images.