The invention discloses a traffic incident detection method under a complex detection condition, which comprises the following steps: acquiring historical data of vehicle position information and lane line position information, and constructing a multi-task learning network model according to a deep learning method; inputting the historical data into a multi-task learning network model for training to obtain a multi-task learning network detection model; extracting image data in real time based on the road monitoring video, and inputting the image data into the multi-task learning network detection model for detection to obtain detection results of vehicles and lane lines; and inputting a detection result into a traffic abnormal event detection algorithm, and detecting an abnormal traffic event on the road. Based on the multi-task learning network model, the problem that the vehicle detection effect is poor under bad conditions is solved, and the same network model detects the vehicle and the lane line at the same time, so that hardware resources of related equipment are saved, the operation efficiency of the model is improved, and the real-time requirement of road video monitoring is met.
本发明公开了一种复杂检测条件下交通事件检测方法,包括:获取车辆位置信息和车道线位置信息的历史数据,根据深度学习方法构建多任务学习网络模型;将历史数据输入多任务学习网络模型进行训练,获得多任务学习网络检测模型;基于道路监控视频实时提取图像数据,将图像数据输入多任务学习网络检测模型进行检测,获得车辆和车道线的检测结果;将检测结果输入交通异常事件检测算法,检测道路上的异常交通事件。本发明基于多任务学习网络模型,解决了在不良条件下车辆检测效果差的问题,由于是同一个网络模型同时检测车辆和车道线,所以不仅节省了相关设备的硬件资源,而且提高了模型的运行效率,满足了公路视频监控实时性的要求。
Traffic incident detection method under complex detection condition
一种复杂检测条件下交通事件检测方法
20.12.2022
Patent
Elektronische Ressource
Chinesisch
Traffic congestion incident detection
TIBKAT | 1996
|Traffic incident detection method based on intelligent spikes
Europäisches Patentamt | 2025
|Microwave-based traffic incident detection method and apparatus
Europäisches Patentamt | 2015
|