基于车道信息融合的车辆行为识别

Vehicle behavior recognition based on lane information fusion

在车辆的自动驾驶和辅助驾驶中,实时分析车辆的运动状态具有重要的实际应用价值。为了实现对车辆行为的判断,提出一种基于车道信息融合的车辆行为识别算法。首先提出一种基于改进Robinson与LSD的模型,运用改进的Robinson算子获取最佳梯度幅值实现对车道的边缘提取,再通过LSD算法实现车道的检测。然后采用一种基于滑动窗口的三次样条插值法对车道进行拟合,最后根据车道参数信息分析车辆的运动状态,结合车辆的中心位置得到车辆的偏离信息。在BDD100K数据集的测试中,本文算法的车道检测准确率为95.61%,车辆行为识别准确率为93.04%,每秒传输帧数达到42.37。实验结果表明,本文算法在不同场景下可以有效地区分车辆的运动状态并给出车辆的偏离信息,具有更高的准确性和鲁棒性。

Real-time analysis of vehicle motion state has important practical application value in automatic driving and assistant driving of vehicles.In order to realize the judgment of vehicle behavior,a vehicle behavior recognition algorithm based on lane information fusion is proposed.Firstly,a model based on improved Robinson and LSD is proposed.The improved Robinson operator is used to obtain the optimal gradient amplitude to realize the edge extraction of the lane,and then the lane detection is realized by LSD algorithm.Then,a cubic spline interpolation method based on sliding window is used to fit the lane.Finally,the motion state of the vehicle is analyzed according to the lane parameter information,and the deviation information of the vehicle is obtained combining with the center position of the vehicle.In the test of BDD100K dataset,the accuracy of lane detection in the algorithm is 95.61%,the accuracy of vehicle behavior recognition is 93.04%,and the number of transmission frames per second reaches 42.37.The experimental results show that the proposed algorithm can effectively distinguish the motion state of the vehicle and give the vehicle deviation information in different scenarios,which has higher accuracy and robustness.