摘要
目前,基于深度学习的目标检测和图像处理的应用范围非常广。在图像处理方面,加入了深度学习能够使一些算法即使在复杂的环境下也能有着不错的识别效果。基于深度学习的方法,通过增加神经网络的网络层数,能够有效的提取出待识别图像的相关特征。其中,逐层设置由浅至深的网络层数就能够提高目标检测的效果。YOLOv3是一种快速的目标检测算法,但其对部分小目标的在检测,识别上不太准确。创新性的通过在基础模型上的改进和增加anchor数量的方法来实现算法的优化。通过实验令该算法在VOC 07数据集上进行验证,实验结果表明与传统未改进的YOLOv3模型相比,在精度上提高了约1.4倍。经过理论阐述和实验成果证明,所提出的方法确实具备更高的执行效果和可靠性,大大提升了模型在数据集上的平均准确率,说明此改进具有有效性和可行性。
Real-time traffic flow prediction based on artificial intelligence refers to the use of large data of vehicles in intelligent transportation system to calculate the depth of artificial intelligence and then predict real-time traffic flow.However,the related work of AI is facing many challenges,such as the large number of traffic physical objects,the low accuracy of spatial location of traffic information and the diversity of vehicles,which will have a great impact on real-time traffic flow forecasting.Therefore,this paper studies the artificial intelligence algorithms of traffic physical object(TPO)and traffic information space(TIS)under the background of intelligent transportation system.A state-based filtering module is used to improve the prediction accuracy of ITS.The theoretical analysis and experimental results show that the proposed method has high execution efficiency and improves the accuracy of real-time traffic flow prediction.
作者
张培培
王昭
王菲
Zhang Peipei;Wang Zhao;Wang Fei(Xi′an Traffic Engineering Institude,Xi′an 710300,China)
出处
《国外电子测量技术》
2020年第8期34-39,共6页
Foreign Electronic Measurement Technology
基金
陕西省教育厅科研计划(18JK1042)项目资助。
