基于注意力机制的车辆异常轨迹检测
Vehicle abnormal trajectory detection based on attention mechanism
柴曼妮1
作者信息
- 1. 太原师范学院计算机科学与技术学院,晋中 030619
- 折叠
摘要
随着城市交通量的增大,安全隐患越来越多,车辆轨迹异常检测对于安全驾驶领域来说也越来越重要.为了更好地提取轨迹的特征,在循环神经网络检测的基础上加入了卷积神经网络,利用卷积加循环的神经网络检测学习轨迹序列信息,并且结合了注意力机制,通过这种技术结合的方法,进一步提高轨迹嵌入的质量.结果表明,该轨迹异常检测方法的性能在各项指标上显著优于其他检测算法,验证了该异常检测方法的有效性和实用性.
Abstract
With the expansion of urban traffic,there are more and more safety hazards,and vehicle track anomaly detection has become more and more important in the field of driving safety.In order to better extract the features of the trajectory,convolu-tional neural network is added on the basis of recurrent neural network detection.The convolutional and cyclic neural network de-tection is used to learn the trajectory sequence information,and the attention mechanism is combined.Through this method,the quality of trajectory embedding is further improved.The results show that the anomaly detection performance of the proposed trajec-tory anomaly detection method is significantly better than other detection algorithms in various indexes,which verifies the effective-ness and practicability of the proposed anomaly detection method.
关键词
车辆轨迹异常检测/循环神经网络/注意力机制Key words
vehicle track anomaly detection/recurrent neural network/attention mechanism引用本文复制引用
出版年
2024
NETL
NSTL
万方数据