时序CNN与稀疏光流融合的视频人车异常检测  

Abnormal Detection in Surveillance Video Based on Fusion Temporal CNN and Sparse Optical Flow

作　　者：李飞[1] 李婷[2] 夏红霞[2,3] 刘文璇 LI Fei;LI Ting;XIA Hongxia;LIU Wenxuan(Hubei Academy of Agricultural Sciences,Wuhan 430064;School of Computer Science and Technology,Wuhan University of Technology,Wuhan 430070;Hubei Key Laboratory of Transportation Internet of Things(Wuhan University of Technology),Wuhan 430070)

机构地区：[1]湖北省农业科学院,武汉430064 [2]武汉理工大学计算机科学与技术学院,武汉430070 [3]交通物联网技术湖北省重点实验室(武汉理工大学),武汉430070

出　　处：《计算机与数字工程》2022年第8期1696-1701,共6页Computer & Digital Engineering

基　　金：中央高校基本科研业务费专项资金(编号:191010001);交通物联网技术湖北省重点实验室开放基金项目(编号:2018IOT003);湖北省自然科学基金重点类创新群体项目(编号:2017CFA012)资助。

摘　　要：大多现存的监控视频异常检测方法都是基于繁琐的特征提取,检测性能不佳,时间效率低下。考虑到视频检测中效率对实际应用的影响,提出一种基于时序CNN与鲁棒稀疏光流融合的视频异常检测方法。该方法在CNN基础上结合语义特征与时间特征的优点定位异常区域,进一步利用光流对异常行为进行检测。实验结果表明,该方法可以有效降低视频检测中的错误率,并且保持较高的准确率以及高效地对视频中的人车异常行为进行检测。Most existing surveillance video anomaly detection methods are based on cumbersome feature extraction,with poor detection performance and low time efficiency. Considering the impact of efficiency on practical applications,a video abnormal detection method based on the fusion of temporal CNN and robust sparse optical flow is proposed. This method is based on CNN to detect the abnormal area with the change of time sequence,and uses optical flow to detect the abnormal area in depth. Experimental results show that this method can effectively reduce the error rate in video detection,maintain a high accuracy rate,and efficiently detect human and vehicle anomalies.

关 键 词：人车异常检测 时序光流融合 时序卷积神经网络 光流法 监控视频 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]