基于环视的道路标线检测与识别

Lane detection and recognition based on around-view system

针对传统道路标线检测系统采用前视摄像头易受车辆遮挡等问题,提出了一种基于环视的多类型道路标线检测与识别方法.该方法利用车辆四周的路面信息,根据近视野道路标线呈直线的特性,采用Hough变换提取环视图像主方向以矫正图像角度,然后建立x方向统计直方图提取行车线区域,对此区域建立y方向统计直方图进一步提取标线特征,通过特征分析识别道路标线的虚实和线数信息,利用HSV色彩分析获取颜色信息.最后分别用卡尔曼滤波器和置信度度量方法对检测和识别结果进行滤波并实现检测.实验结果表明该方法对环视图像中的道路标线有很高的检测率和识别率.

A multi-type lane detection and recognition method using onboard around-view system was proposed,which could provide distinct road information around the vehicle and was less susceptible to occlusion by other vehicles.Firstly,the principle direction of the image was rectified based on Hough transform,as lane markings close to the vehicle are straight.Then xspace feature histogram was used to detect regions of interest and yspace feature histogram was used to recognize the properties of lane markings,such as number of lines,solid/dashed,and so on.Color information was analyzed in the HSV color space.Finally,a Kalman filter and confidence measurement was used to reduce the impact of noise.Experimental results demonstrated the effectiveness of the proposed method.