The invention discloses a night traffic flow detection method based on ShuffleNetv2-YOLOv5s, and the method comprises the steps: firstly obtaining a needed data set; the data set training module is used for training a ShuffleNetv2-YOLOv5s network by a data set; deploying the trained network to a local traffic flow detection platform; video images acquired on the platform are subjected to vehicle judgment by using the network, if the judgment is yes, the video images are marked by using an anchoring frame, and if not, the video images are not processed; and tracking the position change of the vehicle marked by the anchoring frame between continuous frames through a Depsort algorithm, calculating the displacement of the vehicle in each frame according to the tracked vehicle position information, dividing the displacement by a time interval to obtain the average speed of the vehicle, and displaying the average speed on a screen. On the basis of ensuring the detection speed, the size of the network is reduced, and the problem that a high-level vehicle feature information extraction task of a network model is difficult to complete in time due to the conditions of low brightness, high noise, low contrast ratio, complex light source and the like in a night road scene is effectively solved.
本发明公开了一种基于ShuffleNetv2‑YOLOv5s的夜间交通流检测方法,首先获取所需的数据集;用于数据集训练ShuffleNetv2‑YOLOv5s网络;将训练完成的网络部署到本地交通流检测平台;而后对平台上所采集的视频图像使用该网络进行车辆判断,若判断是则用锚定框标记出来,若否不做处理;通过deepsort算法来跟踪锚定框标记出来的车辆在连续帧之间的位置变化,根据跟踪到的车辆位置信息计算出每一帧中车辆的位移,通过位移除以时间间隔得到车辆的平均速度并显示在屏幕上。本发明在保证检测速度的基础上,降低网络的大小,且有效地解决了夜间道路场景中亮度低、噪声大、对比度低、光源来源复杂等条件导致网络模型高层的提取车辆特征信息任务难以及时完成的问题。
Night traffic flow detection method based on ShuffleNetv2-YOLOv5s
基于ShuffleNetv2-YOLOv5s的夜间交通流检测方法
2024-05-28
Patent
Electronic Resource
Chinese
Lightweight Detection of Fasteners with YOLOv8 Combined with ShuffleNetV2
Springer Verlag | 2024
|UAV target detection algorithm based on improved YOLOv5s
SPIE | 2023
|Vehicle Target Detection Algorithm Based on CBAM-YOLOv5s
IEEE | 2024
|