期刊导航

论文摘要

一种基于脉线流卷积神经网络的人群异常行为检测算法

A Crowd Abnormal Behavior Detection Methodbased on Streak Flow CNN

作者:蒋俊(西南石油大学 计算机科学学院);张卓君(西南石油大学 计算机科学学院);高明亮(山东理工大学 电气与电子工程学院);徐立宾(山东理工大学 电气与电子工程学院);潘金凤(山东理工大学 电气与电子工程学院)

Author:JIANG Jun(School of Computer Science,Southwest Petroleum University);ZHANG Zhuojun(School of Computer Science,Southwest Petroleum University);GAO Mingliang(School of Electrical Electronic Engineering,Shandong University of Technology);XU Libin(School of Electrical Electronic Engineering,Shandong University of Technology);Pan Jinfeng(School of Electrical Electronic Engineering,Shandong University of Technology)

收稿日期:2020-02-17          年卷(期)页码:2020,52(6):-

期刊名称:工程科学与技术

Journal Name:Advanced Engineering Sciences

关键字:人群异常检测;脉线流;时空流网络模型;残差网络

Key words:crowd abnormal behavior; streak line flow; spatial-temporal CNN; Resnet101

基金项目:国家自然科学基金: "基于纹线流和群集性的群体异常行为识别和分级预警研究"(61601266),国家自然科学基金:"时变稀疏信号压缩观测的低秩稀疏分解研究"(61801272)

中文摘要

在计算机视觉领域,人群异常行为检测技术可以广泛应用于视频监控、智能视频分析、群体行为识别等领域,因此,受到了学者们的广泛关注。由于视频中人群目标具有尺度变化大、透视形变、标注偏置等特点,人群异常行为检测依然是一个具有挑战性的难题。为此,本文提出了一种基于脉线流和卷积神经网络的人群异常行为检测方法(Streak Flow CNN Abnormal Behavior Detection,简称SFCNN-ABD)。SFCNN-ABD通过卷积神经网络获取显著的人群行为空域特征,并通过脉线流结合卷积神经网络获取人群行为时域特征。SFCNN-ABD是一个双流网络,网络结构由两个深度残差网络作为骨干网络,分别为空域网络和时域网络。其中,空间域网络的输入是原始视频帧,提取人群行为的表观特征,而时域网络利用脉线流提取人群行为的运动特征,脉线流能更准确地识别场景中的空域和时域变化,因而能进一步提升人群异常行为检测的准确性。最后将两个网络的输出进行融合,完成人群异常行为的检测。在UMN和VIF两个公开基准数据集进行了测试,实验结果表明本文方法的性能优于当前主流算法,验证了本文方法的有效性。

英文摘要

Abnormal behavior detection in crowded scenes has drawn extensive interests in computer vision community due to its various applications, e.g., video surveillance, video analytics, and action recognition. However, it is still a challenging task due to the large-scale variation, perspective distortion, and labeling biases. In this paper, an abnormal crowd behavior detection method based on streak flow and convolutional neural network (SFCNN-ABD) is proposed to address this problem. The SFCNN-ABD learns salient spatial features via a convolutional neural network (CNN) and temporal features with the aid of streak flow and CNN. The SFCNN-ABD consists of two stream CNNs, in which the backbone network is composed of two deep residual networks which are spatial network and temporal network respectively. The spatial network utilizes the raw video frames to extract the appearance features of crowded scenes, whereas the temporal network uses the streak flow to extract the kinematic features. The streak flow can accurately recognize the spatial and temporal changes in the scene, and thus can further boost the proposed the performance. Finally, the outputs of these two networks are fused to detect the abnormal crowd behavior. Experimental results on UMN and VIF benchmark datasets indicate that the proposed method outperforms several state-of-the-art methods and thus illustrates the efficiency of the proposed method.

下一条:金沙江白格滑坡裂缝区失稳概率分析

关闭

Copyright © 2020四川大学期刊社 版权所有.

地址:成都市一环路南一段24号

邮编:610065