一种有效的车道线识别与偏道预警方法

An Efficient Method of Lanes Recognition and Departure Warning

室外自然光照条件下进行车道线检测时,光照强度和视野角度变化常常会对识别的结果产生较大影响。为此提出了一种基于自适应视频源参数调解的车道线实时检测方法。该方法通过实时动态调解视频源参数(白平衡、Gamma值)等适应室外光照环境。采用Ada-boost算法训练分类器获得感兴趣车道线区域图像,有效地缩小了图像处理范围,减少了运算量。提出了一种基于冒泡排序法的哈夫变换方法,在边缘检测过程中有效地过滤背景噪声,提高车道线检测的鲁棒性。实验表明视频源参数动态调解算法可以适应各种室内、外自然光照条件,提高了车道线识别精度。车道线中心线和侧向距离偏差检测及预警过程能够及时给出预誓信息,有效地避免了高速行驶的自主车辆偏道现象。

The lane recognition problem with the intensity of illumination and angular field of view in the outside lightness is discussed. A recognition method of lanes is proposed, based on self-adaption of video parameter. The method can adapt to the outdoor natural lightness by adjusting video parameters dynamically, such as white balance, Gamma value, etc. Adaboost algorithm is used to obtain those interesting areas of video images. The method only deals with the partial image of lanes, and reduces computation efficiently. An improved Hough transform method, bosed on bubble-up ranking, is presented to filter the background noises and improve the robustness of recognition during edge detection. The experiments show that the algorithm can reduce the influence on the recognition of lanes under natural lightness to some extent, and improve the recognition precision. Lane departure detection and warning for direction of center line or lateral pose offers the real-time warning information and avoids the departure of direction and pose.