一种智能汽车的实时道路边缘检测算法

以无人驾驶汽车为平台,针对结构化、半结构化道路下无人驾驶汽车道路边缘检测问题,提出了一种智能汽车的实时道路边缘检测算法。该算法首先对获取的激光雷达数据点云进行标定、分层与中值滤波,然后提取各层的左右边界点,而后利用随机抽样一致性算法(简称Ransac)对左右边界点集进行直线拟合,最后用卡尔曼滤波算法进行跟踪,从而实现实时的道路边缘检测。经实验验证,该算法准确率高,可靠性强,能够准确完成道路边缘检测,可以满足实时系统的要求,并已经成功应用于2014年的"智能汽车未来挑战赛",而且取得了第三名的好成绩。

A Real-time Road Boundary Detection Algorithm Based on Driverless Cars

Pointing to road boundary detection problem of driverless cars in structured and semi-structured road, the paper proposed a new real-time road boundary detection algorithm based on driverless cars as a platform. The algorithm firstly includes the following steps:calibration, layering and median filtering will be made according to the obtained lidar point cloud data. And then the left and right road boundary point of each layer will be extracted. The extracted road boundary points are then straight line fitted using Ransac algorithm. Finally, straight line is tracked using Kalman filtering, thus the real-time road boundary detection is achieved. The test results show that this boundary detection algorithm is with high accuracy and reliability and is able to accurately accomplish the boundary detection task, which can satisfy the requirements of real-time system. And this algorithm has been applied successfully in "The Future Challenge Competition for Driverless Cars" 2014 and obtained the third place.