一种改进R(2+1)D网络的暴力行为检测方法

doi:10.19665/j.issn1001-2400.2022.02.018

Abstract

Abstract:

In public security,the complex violence behavior detection is of important research value.Traditional methods are mainly based on hand-crafted features,but they have a limited generalization ability.The existing deep learning network models have a better tolerance,but those kinds of methods face the challenge of obtaining a high accuracy.To solve the above problems,a novel violent behavior detection method is proposed in this paper by combining an improved R (2+1) D network and the dense connection idea.In the branch of the traditional R (2+1) D residual module,the three-quarters of the feature map is ignored due to the convolution operation with strides of 2.In this paper,the convolution operation is optimized for a pooling operation and a convolution operation with strides of 1,and the detection accuracy is increased by 2.3%.Besides,the dense connection idea is adopted into the residual module to establish the connection between different convolutional layers.The improvement could alleviate the problem of gradient dissipation during the training process,and the detection accuracy is further improved by 1.46%.Experimental results on one public dataset and one self-built dataset demonstrate the effectiveness of the proposed method for the complex violence behavior detection.

Key words: violence detection, (2+1) D dense residual block, residual network, deep learning

CLC Number:

TP391.4

WANG Yong,JIN Weizhao,FENG Wei,QUAN Yinghui. Improved violent behavior detection method for the R(2+1)D network[J].Journal of Xidian University, 2022, 49(2): 155-163.

Figures/Tables 16

网络层	输出特征尺寸	R(2+1)D网络结构参数	(2+1)D密集残差网络结构参数
卷积模块1	16×56×56	$1 × 7 × 7,45,1 × 2 × 2 3 × 1 × 1,64,1 × 1 × 1$
卷积模块2	16×56×56	$1 × 3 × 3,144,1 × 1 × 1 3 × 1 × 1,64,1 × 1 × 1 1 × 3 × 3,144,1 × 1 × 1 3 × 1 × 1,64,1 × 1 × 1$ ×2	$1 × 3 × 3,144,1 × 1 × 1 3 × 1 × 1,64,1 × 1 × 1 1 × 3 × 3,144,1 × 1 × 1 3 × 1 × 1,64,1 × 1 × 1 1 × 1 × 1,64,1 × 1 × 1$ ×1
卷积模块3	8×28×28	$1 × 3 × 3,230,1 × 2 × 2 3 × 1 × 1,128,2 × 1 × 1 1 × 3 × 3,230,1 × 1 × 1 3 × 1 × 1,128,2 × 1 × 1$ ×2	$1 × 3 × 3,230,1 × 2 × 2 3 × 1 × 1,128,2 × 1 × 1 1 × 3 × 3,230,1 × 1 × 1 3 × 1 × 1,128,2 × 1 × 1 1 × 1 × 1,128,1 × 1 × 1$ ×1
卷积模块4	4×14×14	$1 × 3 × 3,460,1 × 2 × 2 3 × 1 × 1,256,2 × 1 × 1 1 × 3 × 3,460,1 × 1 × 1 3 × 1 × 1,256,2 × 1 × 1$ ×2	$1 × 3 × 3,460,1 × 2 × 2 3 × 1 × 1,256,2 × 1 × 1 1 × 3 × 3,460,1 × 1 × 1 3 × 1 × 1,256,2 × 1 × 1 1 × 1 × 1,256,1 × 1 × 1$ ×1
卷积模块5	2×7×7	$1 × 3 × 3,921,1 × 2 × 2 3 × 1 × 1,512,2 × 1 × 1 1 × 3 × 3,921,1 × 1 × 1 3 × 1 × 1,512,2 × 1 × 1$ ×2	$1 × 3 × 3,921,1 × 2 × 2 3 × 1 × 1,512,2 × 1 × 1 1 × 3 × 3,921,1 × 1 × 1 3 × 1 × 1,512,2 × 1 × 1 1 × 1 × 1,512,1 × 1 × 1$ ×1
	1×1×1	Global Average Pooling,FClayer,Softmax

References 14

[1]	GIANNAKOPOULOS T, KOSMOPOULOS D, ARISTIDOU A, et al. Violence Content Classification Using Audio Features[C]// Hellenic Conference on Artificial Intelligence.Heidelberg:Springer, 2006:502-507.
[2]	LIN J, WANG W. Weakly-Supervised Violence Detection in Movies with Audio and Video Based Co-Training[C]// Pacific-Rim Conference on Multimedia.Heidelberg:Springer, 2009:930-935.
[3]	ZHANG T, YANG Z, JIA W, et al. A New Method for Violence Detection in Surveillance Scenes[J]. Multimedia Tools and Applications, 2016, 75(12):7327-7349. doi: 10.1007/s11042-015-2648-8
[4]	SENST T, EISELEIN V, KUHN A, et al. Crowd Violence Detection Using Global Motion-Compensated Lagrangian Features and Scale-Sensitive Video-Level Representation[J]. IEEE Transactions on Information Forensics and Security, 2017, 12(12):2945-2956. doi: 10.1109/TIFS.2017.2725820
[5]	ZHOU P, DING Q, LUO H, et al. Violence Detection in Surveillance Video Using Low-Level Features[J]. PLoS ONE, 2018, 13(10):0203668.
[6]	SAMUEL D J R, FNIL E, MANOGARAN G, et al. Real Time Violence Detection Framework for Football Stadium Comprising of Big Data Analysis and Deep Learning through Bidirectional LSTM[J]. Computer Networks, 2019, 151:191-200. doi: 10.1016/j.comnet.2019.01.028
[7]	MAHMOODI J, SALAJEGHE A. A Classification Method Based on Optical Flow for Violence Detection[J]. Expert Systems with Applications, 2019, 127:121-127. doi: 10.1016/j.eswa.2019.02.032
[8]	DEEPAK K, VIGNESH L K P, CHANDRAKALA S. Autocorrelation of Gradients Based Violence Detection in Surveillance Videos[J]. ICT Express, 2020, 6(3):155-159. doi: 10.1016/j.icte.2020.04.014
[9]	DONG Z, QIN J, WANG Y. Multi-Stream Deep Networks for Person to Person Violence Detection in Videos[C]// Chinese Conference on Pattern Recognition.Heidelberg:Springer, 2016:517-531.
[10]	ZHOU P, DING Q, LUO H, et al. Violent Interaction Detection in Video Based on Deep Learning[J]. Journal of Physics:Conference Series, 2017, 844:012044. doi: 10.1088/1742-6596/844/1/012044
[11]	LI H, WANG J, HAN J, et al. A Novel Multi-Stream Method for Violent Interaction Detection Using Deep Learning[J]. Measurement and Control, 2020, 53(5-6):796-806. doi: 10.1177/0020294020902788
[12]	周智, 朱明, KHAN Y. 基于3D-CNN的暴力行为检测[J]. 计算机系统应用, 2017, 26(12):207-211.
	ZHOU Zhi, ZHU Ming, KHAN Yahya. Violence Behavior Detection Based on 3D-CNN[J]. Computer Systems & Applications, 2017, 26(12):207-211.
[13]	TRAN D, WANG H, TORRESANI L, et al. A Closer Look at Spatiotemporal Convolutions for Action Recognition[C]// 2018 IEEE Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE, 2018:6450-6459.
[14]	HE T, ZHANG Z, ZHANG H, et al. Bag of Tricks for Image Classification with Convolutional Neural Networks[C]// 2019 IEEE Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE, 2019:558-567.

网络模型	参数量	计算量
3D-CNN^[12]	20.96×10⁶	37.70×10⁹
R(2+1)D网络	33.18×10⁶	41.77×10⁹
改进后R(2+1)D网络	14.36×10⁶	20.82×10⁹
(2+1)D密集残差网络	15.06×10⁶	21.60×10⁹

方法	曲棍球比赛数据集精度/%	自制数据集精度/%
3D-CNN^[12]	93.8^[12]	92.96±0.19
R(2+1)D^[13]	92.71±0.39	92.61±0.62
改进后R(2+1)D	95.01±0.37	93.60±0.47
(2+1)D密集残差网络	96.48±0.38	94.57±0.23

[1]	GAO Jie,HUO Zhiyong. Algorithmfor image inpainting in generative adversarial networks based on gated convolution [J]. Journal of Xidian University, 2022, 49(1): 216-224.
[2]	ZHANG Yan,WANG Xiangyu,ZHANG Zhongwei,SUN Yemei,LIU Shudong. Boundary-aware network for building extraction from remote sensing images [J]. Journal of Xidian University, 2022, 49(1): 236-244.
[3]	LIU Jiawei,ZHANG Wenhui,KOU Xiaoli,LI Yanni. Harnessing adversarial examples via input denoising and hidden information restoring [J]. Journal of Xidian University, 2021, 48(6): 23-31.
[4]	SONG Jianfeng,MIAO Qiguang,WANG Chongxiao,XU Hao,YANG Jin. Multi-scale single object tracking based on the attention mechanism [J]. Journal of Xidian University, 2021, 48(5): 110-116.
[5]	LI Peng,FENG Cunqian,XU Xuguang,TANG Zixiang. Ballistic target fretting classification network based on Bayesian optimization [J]. Journal of Xidian University, 2021, 48(5): 139-148.
[6]	ZHANG Yuhao,CHENG Peitao,ZHANG Shuhao,WANG Xiumei. Lightweight image super-resolution with the adaptive weight learning network [J]. Journal of Xidian University, 2021, 48(5): 15-22.
[7]	YAN Jia,CAO Yudong,REN Jiaxing,CHEN Donghao,LI Xiaohui. Deep asymmetric compression Hashing algorithm [J]. Journal of Xidian University, 2021, 48(5): 212-221.
[8]	NING Yang,DU Jianchao,HAN Shuo,YANG Chuankai. Fire segmentation based on the improved DeeplabV3+ and the analytical method for fire development [J]. Journal of Xidian University, 2021, 48(5): 38-46.
[9]	ZHOU Peng,YANG Jun. Semantic segmentation of remote sensing images based on neural architecture search [J]. Journal of Xidian University, 2021, 48(5): 47-57.
[10]	QI Yanjun,KONG Yueping,WANG Jiajing,ZHU Xudong. Gait recognition method combining LSTM and CNN [J]. Journal of Xidian University, 2021, 48(5): 78-85.
[11]	ZHAN Keyu,SUN Yue,LI Ying. Video super-resolution based on multi-scale 3D convolution [J]. Journal of Xidian University, 2021, 48(5): 8-14.
[12]	HUI Haisheng,ZHANG Xueying,WU Zelin,LI Fenglian. Method for stroke lesion segmentation using the primary-auxiliary path attention compensation network [J]. Journal of Xidian University, 2021, 48(4): 200-208.
[13]	SUN Haojie,LI Miaoyu,ZHANG Panpan,XU Pengfei. Self-supervised facial asymmetry learning for automatic evaluation of facial paralysis [J]. Journal of Xidian University, 2021, 48(3): 115-122.
[14]	ZHANG Hua,GAO Haoran,YANG Xingguo,LI Wenmin,GAO Fei,WEN Qiaoyan. TargetedFool:an algorithm for achieving targeted attacks [J]. Journal of Xidian University, 2021, 48(1): 149-159.
[15]	YANG Hongyu,ZENG Renyun. Method for assessment of network security situation with deep learning [J]. Journal of Xidian University, 2021, 48(1): 183-190.

Improved violent behavior detection method for the R(2+1)D network

RichHTML

PDF (PC)

Like

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 16

References 14

Related Articles 15

Metrics

Comments

Recommended 10