Human Detection of Damage Behavior for Vending Cabinets Based on Improved YOLOv4-Tiny

doi:10.3778/j.issn.1002-8331.2212-0057

Abstract

Abstract: The safety inspection of unmanned containers has always been a hot topic in the retail field. Aiming at the problem that the existing manual monitoring cannot timely and effectively capture the damage behavior of some consumers to the self-service vending cabinet and its internal products, an improved YOLOv4-Tiny oriented human detection method of damage behavior for vending cabinets is proposed. First of all, the surveillance video collected in the real scene is preprocessed, and the production of the data set DMGE-Act is completed to solve the problem of insufficient scene image data sources. Then, an improved model based on YOLOv4-Tiny, YOLOv4-TinyX, is proposed. By modifying the activation function of the neural network for smooth approximation, CBAM is introduced after the largest feature extraction layer of the backbone feature extraction network, and in strengthening the feature extraction network. After the upsampling operation layer, two different attention mechanism modules of CA are introduced, and the data imbalance is corrected, which effectively improves the feature extraction and detection capabilities of the algorithm. Through comparative experimental analysis, the improved model parameters are only increased by 2×104, while the average precision mAP is increased by 10.29 percentage points. The results show that the algorithm remains lightweight and the detection accuracy of damage behavior is significantly improved.

Key words: unattended, damage behavior, YOLOv4-Tiny, smooth approximation, attention mechanism, lightweight

摘要： 无人货柜的安全检测一直是零售领域的热点话题。针对现有人工监控无法及时且有效地捕捉到部分消费者对自助售货柜及其内部商品的损坏行为这一问题，提出了一种改进YOLOv4-Tiny的面向售货柜损害行为人体检测方法。将真实场景采集到的监控视频进行预处理，完成对数据集DMGE-Act的制作，解决场景图像数据源不足的问题。提出了基于YOLOv4-Tiny的改进模型——YOLOv4-TinyX，通过修改神经网络的激活函数进行平滑逼近，分别在主干特征提取网络的最大特征提取层后引入CBAM，在加强特征提取网络中的上采样操作层后引入CA两种不同的注意力机制模块，并且进行了数据不平衡的修正，有效提升了算法的特征提取与检测能力。通过对比实验分析，改进后的模型参数量仅增加2×104的同时，平均精度均值mAP提升了10.29个百分点，结果表明该算法保持轻量化且对损害行为的检测精度有显著提升。

关键词: 无人值守, 损害行为, YOLOv4-Tiny, 平滑逼近, 注意力机制, 轻量化

YIN Min, JIA Xinchun, ZHANG Xueli, FENG Jiangtao, FAN Xiaoyu. Human Detection of Damage Behavior for Vending Cabinets Based on Improved YOLOv4-Tiny[J]. Computer Engineering and Applications, 2024, 60(8): 234-241.

殷民, 贾新春, 张学立, 冯江涛, 范晓宇. 改进YOLOv4-Tiny的面向售货柜损害行为人体检测[J]. 计算机工程与应用, 2024, 60(8): 234-241.

References

[1] KARPATHY A, TODERICI G, SHETTY S, et al. Large-scale video classification with convolutional neural networks[C]//Proceedings of the Computer Vision & Pattern Recognition, 2014: 1725-1732.
[2] SIMONYAN K, ZISSERMAN A. Two-stream convolutional networks for action recognition in videos[C]//Advances in Neural Information Processing Systems, 2014.
[3] DONAHUE J, HENDRICKS L A, GUADARRAMA S, et al. Long-term recurrent convolutional networks for visual recognition and description[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2015: 2625-2634.
[4] NG Y H, HAUSKNECHT M, VIJAYANARASIMHAN S, et al. Beyond short snippets: deep networks for video classification[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2015: 2625-2634.
[5] CAO Z, HIDALGO G, TOMAS S, et al. OpenPose: realtime multi-person 2D pose estimation using part affinity fields[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 43(1): 172-186.
[6] BALDERAS D, PONCE P, MOLINA A . Convolutional long short term memory deep neural networks for image sequence prediction[J]. Expert Systems with Application, 2019, 122: 152-162.
[7] ZHANG F, WANG Y, ZHANG Z. View-invariant action recognition in surveillance videos[C]//Proceedings of the First Asian Conference on Pattern Recognition, 2011: 580-583.
[8] LI K L, HUANG H K, TIAN S F, et al. Improving one-class SVM for anomaly detection[C]//Proceedings of the 2003 International Conference on Machine Learning and Cybernetics, 2003: 3077-3081.
[9] PENG X, WANG L, WANG X, et al. Bag of visual words and fusion methods for action recognition: comprehensive study and good practice[J]. Computer Vision and Image Understanding, 2016, 150: 109-125.
[10] 余兴.基于深度学习的视频行为识别技术研究[D].成都：电子科技大学, 2018.
YU X. Research on video behavior recognition technology based on deep learning[D].Chengdu: University of Electronic Science and Technology of China, 2018.
[11] 张瑞, 李其申, 储珺.基于3D卷积神经网络的人体动作识别算法[J]. 计算机工程, 2019, 45(1): 259-263.
ZHANG R, LI Q S, CHU J. Human action recognition algorithm based on 3D convolution neural network[J]. Computer Engineering, 2019, 45(1): 259-263.
[12] 钱国华, 程芳芳, 朱孝慈, 等.电梯内异常行为检测系统设计[J].工业控制计算机, 2019, 32(11): 91-92.
QIAN G H, CHENG F F, ZHU X C, et al. Design of abnormal behavior detection system in elevator[J]. Industrial Control Computer, 2019, 32(11): 91-92.
[13] CAO Y, XU H, YANG Q. Computer-vision-based abnormal human behavior detection and analysis in electric power plant[C]//Proceedings of the 2021 33rd Chinese Control and Decision Conference, 2021: 1578-1583.
[14] 赵凤, 李永恒, 李晶, 等.基于改进YOLOv4-tiny的轻量化室内人员目标检测算法[J].电子与信息学报, 2022, 44(11): 3815-3824.
ZHAO F, LI Y H, LI J, et al. Lightweight indoor personnel target detection algorithm based on improved YOLOv4-tiny[J]. Journal of Electronics and Information, 2022, 44(11): 3815-3824.
[15] 程浩然, 王薪陶, 李俊燃, 等.改进YOLOv4-tiny的疫情协同口罩佩戴检测方法[J].计算机工程与应用, 2023, 59(20): 208-218.
CHENG H R, WANG X T, LI J R, et al. Improving YOLOv4-tiny’s epidemic collaborative mask wearing detection method[J].Computer Engineering and Applications, 2023, 59(20): 208-218.
[16] 徐长友, 樊绍胜, 朱航.采用通道域注意力机制Deeplabv3+算法的遥感影像语义分割[J].控制工程, 2023, 30(2): 368-375.
XU C Y, FAN S S, ZHU H. Remote sensing image semantic segmentation using channel domain attention mechanism Deeplabv3+ algorithm[J]. Control Engineering, 2023, 30(2): 368-375.
[17] WOO S, PARK J, LEE J Y, et al. CBAM: convolutional block attention module[C]//Proceedings of the European Conference on Computer Vision, 2018: 3-19.
[18] 张廓, 陈章进, 张岩.改进YOLOv4-Tiny的SAR图像目标快速检测方法[J].计算机工程与应用, 2023, 59(14): 209-216.
ZHANG K, CHEN Z J, ZHANG Y. Improved YOLOv4-Tiny method for fast target detection in SAR images[J]. Computer Engineering and Applications, 2023, 59(14): 209-216.
[19] 王晓雯, 梁博, 刘芳芳.基于注意力机制与加权盒函数的YOLOv5的行人摔倒检测算法[J].山西大学学报(自然科学版), 2023, 46(2): 334-341.
WANG X W, LIANG B, LIU F F. Pedestrian fall detection algorithm based on attention mechanism and weighted box function of YOLOv5[J]. Journal of Shanxi University (Natural Science Edition), 2023, 46(2): 334-341.
[20] HOU Q, ZHOU D, FENG J. Coordinate attention for efficient mobile network design[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021: 13713-13722.
[21] 王长清, 贺坤宇, 蒋帅.改进YOLOv4-tiny网络的狭小空间目标检测方法[J].计算机工程与应用, 2022, 58(10): 240-248.
WANG C Q, HE K Y, JIANG S. An improved small space target detection method for YOLOv4-tiny network[J]. Computer Engineering and Applications, 2022, 58(10): 240-248.
[22] LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection[C]//Proceedings of the IEEE International Conference on Computer Vision, 2017: 2980-2988.