摘要
随着网点数量的快速增长,物流快递驿站的安全问题得到了越来越多的重视。文章提出了一种基于深度学习的物流快递驿站异常行为识别方法,其采用卷积神经网络ResNet152提取视频的每一帧图像特征,通过递归神经网络LSTM对帧与帧之间的时序关系进行建模,并采用多路分支网络架构以适应不同物流快递驿站的摄像机视角变化。在实际物流快递驿站场景下的实验结果验证了本文方法的有效性。
With the rapid growth of the number of logistics courier stations,the security issues of which have received more and more attention.This paper presents a method of abnormal behavior recognition for logistics courier stations based on deep learning,which adopts the convolutional neural network ResNet152 to extract the image features of each frame in the video,and uses the recurrent neural network LSTM to model the temporal relationship between frames.Moreover,it proposes a multi-branch network architecture to adapt to the change of camera viewpoint of different logistics courier stations.The experimental results in the scenario of actual logistics courier station verify the effectiveness of the proposed method.
作者
陈松乐
孙知信
CHEN Songle;SUN Zhixin(Engineering Research Center of Post Big Data Technology and Application of Jiangsu Province,Nanjing University of Posts and Telecommunications,Nanjing 210003,China;Research and Development Center of Post Industry Technology of the State Posts Bureau(Internet of Things Technology),Nanjing University of Posts and Telecommunications,Nanjing 210003,China)
出处
《物流科技》
2020年第9期50-54,共5页
Logistics Sci-Tech
基金
国家自然科学基金项目(61672299、61972208)
江苏省高校自然科学基金项目(18BC051)。
关键词
物流快递驿站
行为识别
卷积神经网络
递归神经网络
logistics courier station
behavior recognition
convolutional neural network
recurrent neural network
