基于YOLOv8的自动驾驶车道检测改进算法

Enhance Curve Lane Detection Algorithm for Autonomous Vehicles based on YOLOv8

摘要:

针对自动驾驶领域的车道自动检测中存在的检测准确率低、实际应用难等问题，研究基于YOLO与传统图像处理算法混合的车道检测算法。基于车载传感器拍摄的视频，利用YOLOv8算法检测并标记车前/侧方附近的物体，并将图像视角转换到鸟瞰视角，利用基于滑动窗口的二次多项式法识别当前帧的车道线，融合前序帧的车道信息，检测出当前帧的车道。经过数据集和实际场景的测试表明，算法的检测准确性提升10%以上，检测速度明显提升。

Abstract:

Aiming at the problems of low detection accuracy and difficult practical application in automatic lane detection in the field of automatic driving, the lane detection algorithm based on the mixture of YOLO and traditional image processing algorithms is studied. Based on the video captured by the on-board sensors, the YOLOv8 algorithm is used to detect and mark the objects near the front/side of the vehicle, and the image viewpoint is converted to a bird's-eye view, and the sliding-window-based quadratic polynomial method is used to identify the lane lines in the current frame, and fusing the lane information of the preceeding frames to detect the lane in the current frame. Tests on datasets and real-world scenarios show that the algorithm's detection accuracy is improved by more than 10% and the detection speed is significantly increased.