基于单目视觉的车道偏离预警算法研究

Research oflane departure warning algorithm based on monocular vision

随着科技的发展,智能交通已成为当前的研究热点,自适应巡航是智能交通主要应用之一,而车道偏离预警在主动安全、自适应巡航中起着关键性的作用。因此,提出一种基于视频的车道偏离预警算法。把检测到的车道线的ROI分割为两个子区域,两个子区域分别应用霍夫变换寻找左右车道线。分割的方法可以提高车道线检测的计算速度。综合车辆驾驶状态、驾驶环境等相关影响因素,基于混合高斯隐型马尔科夫模型建立了驾驶人换道意图识别模型,以提高驾驶员换道意图辨别的准确率与灵敏度。

With the development of science and technology,intelligent transportation has become a current research hot spot. Adaptive cruise is one of the main applications of intelligent transportation,and lane departure warning plays a key role in active safety and adaptive cruise. This paper proposes a method of lane departure warning based on video. The ROI of the detected lane lines is divided into two sub regions. The two sub regions are applied to find the left and right lane line by using Hough transform respectively. The segmentation method can improve the speed of the lane detection. Considering vehicle driving condition,driving conditions and related influencing factors comprehensively,the paper established model for driver lane-changing intention recognition based on the Gaussian mixture hidden Markov models to improve the accuracy and sensitivity of the intention identification.