摘要
针对传统LSTM分类模型对车辆直行等行为识别准确率不高的问题,提出一种改进LSTM分类模型。在改进模型中,首先把输入特征进行横向合并,再输入1个LSTM细胞。该模型可以充分利用输入信息,减少计算量,单个LSTM细胞模型具有较强的抗干扰能力、更好的分类效果以及更快的训练速度。实验表明,改进后的模型较改进前总体识别准确率提高1.6%,其中直行识别准确率提高2.04%,训练时间减少3.96 s,识别准确率和训练速度较改进前的模型均有所提升。
In order to solve the problem that the traditional LSTM classification model is not accurate enough to recognize the vehicle behavior such as straight going,an improved LSTM classification model is proposed in this paper.In the improved model,the input features are first combined horizontally,and then one LSTM cell is input.The model can make full use of the input information and reduce the amount of calculation.The single LSTM cell model has strong anti-interference ability,better classification effect and faster training speed.The experimental results show that the improved model improves the overall recognition accuracy by 1.6%,in which the straight line recognition accuracy is increased by 2.04%,the training time is reduced by 3.96 s,and the recognition accuracy and training speed are significantly improved compared with the pre-improved model.
作者
肖海鹏
王任栋
曹波
李炯
XIAO Haipeng;WANG Rendong;CAO Bo;LI Jiong(Fifth Team of Cadets,Army Military Transportation University,Tianjin 300161,China;Institute of Military Transportation,Army Military Transportation University,Tianjin 300161,China;Zhenjiang Campus,Army Military Transportation University,Zhenjiang 212003,China;Unit 95848,Xiaogan 432000,China)
出处
《军事交通学院学报》
2020年第7期83-88,共6页
Journal of Military Transportation University
基金
国家自然科学基金(91220301)
国家重点研发计划(2016YFB0100903)
关键词
车辆行为识别
LSTM分类模型
运动轨迹预测
vehicle behaviour recognition
LSTM classification model
trajectory prediction
