The invention discloses a vehicle identification statistical method based on a multi-attention mechanism network. The method comprises the following steps: acquiring a real-time video of a road which is aerial-photographed by an unmanned aerial vehicle and needs traffic flow statistics; a stepped multi-attention network LMNet designed based on YOLOv5 and HRNet is adopted as a detector, a vehicle target in the real-time video is detected, and position information of a vehicle in the vehicle target is obtained; utilizing an improved Strong SORT network to track the position of the corresponding vehicle according to the obtained position information of the vehicle, so as to obtain the motion trail of the corresponding vehicle; and delimiting a virtual counting line in a road scene aerially photographed by the unmanned aerial vehicle, and counting the number of vehicles according to the condition that the obtained vehicle motion trail passes through the virtual counting line to obtain a traffic flow statistical result on a corresponding road. According to the invention, the detection accuracy is effectively improved, the false detection rate and the missing detection rate are reduced, and the tracking effect is improved.
本发明公开了一种基于多注意力机制网络的车辆识别统计方法,包括:获取无人机航拍的需要统计车流量的道路的实时视频;采用基于YOLOv5和HRNet设计的阶梯式多注意力网络LMNet作为检测器,对实时视频中的车辆目标进行检测,获取其中车辆的位置信息;利用改进的Strong SORT网络,依据所获取的车辆的位置信息对相应的车辆进行位置跟踪,从而获取相应车辆的运动轨迹;在无人机航拍的道路场景中划定虚拟计数线,根据所获取的车辆运动轨迹通过虚拟计数线的情况,统计车辆的数目,得到相应道路上的车流量统计结果。本发明有效提高了检测准确率,降低了误检率和漏检率,提升了追踪效果。
Vehicle identification statistical method based on multi-attention mechanism network
一种基于多注意力机制网络的车辆识别统计方法
2023-01-31
Patent
Electronic Resource
Chinese
Vehicle Detection Based on Deep Neural Network Combined with Radar Attention Mechanism
British Library Conference Proceedings | 2020
|Vehicle trajectory prediction method based on hierarchical attention mechanism
European Patent Office | 2024
|Vehicle Detection Based on Deep Neural Network Combined with Radar Attention Mechanism
SAE Technical Papers | 2020
|