基于双重注意力机制、轨迹预测与无人机遥感技术的人车流量检测方法OA

Accurate detection and statistics of human and vehicle traffic flow are of great significance for public safety and resource manage-ment.A pedestrian and vehicle flow detection method based on dual attention mechanism,trajectory prediction,and drone remote sensing technology is proposed to address the issues of high equipment investment,high maintenance costs,limited detection areas,and susceptibility to environmental factors in existing detection and statistical methods.This method is improved on the basis of the RefineDet network by intro-ducing a dual attention mechanism and replacing the TCB module with feature fusion RNN to enhance the recognition ability of small targets.At the same time,adaptive Kalman filtering is used to predict trajectories,achieving real-time tracking of moving targets,avoiding erroneous statistics caused by target distance and special action modes.Train network models with self built datasets and conduct tests in different scenar-ios.The experimental results show that the proposed improved model has a MOTA improvement of 2.3%to 3.3%compared to contrast models.The detection speed is faster while the accuracy is close to the latest model,basically meeting the real-time requirements;The comprehensive evaluation index F value can reach over 95%in various scenarios,and the false detection rate in various actual scene detection is less than 0.3%.