Abstract: In allusion to the characteristics of unstructured road，this paper puts forward a kind of the edge detection algorithm based on color difference.The algorithm converts the original image’s RGB color space to L*a*b* color space.And using L*a*b* color space lightness and color separation features，it extracts the image edge of road via the color difference information fusion and expels from the effect of a large area of shadow or watermarks on border of the true path.This method raises effectively the accuracy of the lane recognition.Finally，this paper extracts the edge of the road through the Hough transform.Indicated by experiment，this method has many merits，which can be not sensitive to the lane shape，need less priori knowledge and filter best the interference of watermarks or shadow on the road.

Key words: color difference, edge detection, lane recognition

摘要： 针对非结构化道路的特点，提出了一种基于色度差的边缘检测算法。算法将原始图像的RGB颜色空间转化为L*a*b*颜色空间，利用L*a*b*颜色空间中明度与色度分离的特点，通过对色度差的信息融合提取图像道路边缘，排除了大面积阴影、水渍对真实道路边界的干扰，有效地提高了道路识别的准确性。通过Hough变换提取出图像边缘得到道路边缘信息。实验表明，该算法的优点在于对道路形状不敏感，需要先验知识少，能够很好地过滤道路中存在的水渍、阴影等干扰。

关键词: 色度差, 边缘检测, 道路识别