一种基于机器学习的ADAS车道类型判别方法

The Lane Type Identification Method of ADAS Based on Machine Learning

高级汽车辅助驾驶系统(Advanced Driving Assistance System,ADAS)是利用安装在车上的各种传感器,在汽车行驶过程中随时感应周围的环境,收集数据,进行系统的运算与分析,有效增强汽车驾驶的舒适性和安全性。ADAS最重要的功能包括LDW、FCW、BSD、PD、TSR等。目前,应用最广泛的传感器是雷达和摄像头。用单目摄像头进行车道线的识别目前已经有很多解决方案,但是还需要有效的车道线类型的检测方法为自动驾驶过程中的变道决策提供依据。本文提出一种基于机器学习的判断车道线类型的方法,利用车道线相邻区域的直方图特征,有效地解决了车道线类型的判别问题,实验数据表明此方法能够获得99.99%的正确识别率。

The advanced driving assistance system （ ADAS ） uses sensors to collect environment data during driving process, and then conducts analysis to effectively increase the driving comfortability and security. Main functions of ADAS includes LDW, FCW, BSD, PD, TSR, etc. Currently, the most widely used sensors are radars and cameras. There are already many solutions to recognize road line using monocular camera, but effective method to identify the lane type is still needed to help decision-making for automatic driving. This article proposes a lane type identifying method based on machine learning, which uses histogram characteristics of neighboring lane area to effectively recognize lane type. Test data indicates that this method can achieve an accuracy of 99.99%.