基于航拍视频的车辆速度估算研究
Research of Vehicle Speed Estimation Based on UAV Videos

随着新型技术的迅速发展，基于视频的车辆速度识别引起了研究人员的广泛关注。然而，在速度采集过程中仍然存在许多挑战，包括车辆识别精度低、跟踪效果差、镜头抖动等。本文提出了一种车辆速度检测方法。首先，构建了一个基于无人机特定飞行高度下的数据集，并通过深度学习网络训练获得权重文件。然后利用YOLO检测算法、Deep-sort跟踪算法获得车辆的运动轨迹。在此基础上提出了一种基于线性回归思想的车辆速度计算方法。试验表明，本方法具有较高的精度和抗干扰能力。
With the rapid development of new technology, vehicle speed estimation based on video has at-tracted wide attention from researchers. However, there are still many challenges in extracting speed information from UAV videos, including the low accuracy of vehicle identification, the poor tracking effect, the lens jitter. In this paper, we propose a method for vehicle speed estimation. Firstly, we build a traffic dataset based on UAV for specific height, and obtain weight file through the deep-learning algorithm, and then obtain the trajectory through YOLO algorithm and Deep-sort algorithm. Thereafter, we propose a method to estimation vehicle speed based on linear regression algorithm. The result of experiments shows that the strategy has stronger anti-jamming ability and higher precision.