基于YOLO神经网络的垃圾检测与分类

doi:10.16180/j.cnki.issn1007-7820.2022.10.008

摘要/Abstract

摘要：

针对人工分拣垃圾效率低、任务重和环境恶劣等问题,文中提出了基于YOLO的目标检测方法来实现垃圾检测与分类。通过制作特定数据集,使用K-means聚类算法以及Mish激活函数对模型进行调整。根据卷积神经网络的特性,通过在YOLO模型的每个检测头前嵌入CBAM注意力模块,结合PANet增强特征集成能力来提升小目标检测的精度。实验结果表明,文中提出的垃圾检测与分类方法能够准确快速地识别垃圾。相较于YOLOv4,文中所提模型在垃圾数据集上的map值提升了2.81%,其中Cans的识别精度可达94.56%,PlasticBottle的精度提升了6.36%。

关键词: 垃圾识别, 分类, 神经网络, 注意力机制, 深度学习, 数据集, 特征集成, 目标检测

Abstract:

In view of the problems of low efficiency of manual garbage sorting, heavy tasks and harsh environment, this study proposes a YOLO-based target detection method to realize garbage detection and classification. The model is adjusted through making a specific dataset, using K-means clustering algorithm and Mish activation function. According to the characteristics of the convolutional neural network, the CBAM attention module is embedded in front of each detection head of the YOLO model, combined with PANet to enhance the feature integration ability to improve the accuracy of small target detection. The experimental results show that the garbage detection and classification method proposed in this study can accurately and quickly identify garbage. Compared with YOLOv4, the map value of the proposed model on the garbage data set has increased by 2.81%. The recognition accuracy of Cans can reach 94.56%, and the accuracy of PlasticBottle has increased by 6.36%.

Key words: garbage recognition, classification, neural network, attentional mechanism, deep learning, data set, feature integration, object detection

中图分类号:

TP183

张伟,刘娜,江洋,李清都. 基于YOLO神经网络的垃圾检测与分类[J]. 电子科技, 2022, 35(10): 45-50.

ZHANG Wei,LIU Na,JIANG Yang,LI Qingdu. Garbage Detection and Classification Based on YOLO Neural Network[J]. Electronic Science and Technology, 2022, 35(10): 45-50.

图/表 11

图1

图2

图3

图4

图5

图6

图7

表1

表2

表3

图8

参考文献 18

[1]	魏潇潇, 王小铭, 李蕾, 等. 1979-2015年中国城市生活垃圾产生和处理时空特征[J]. 中国环境科学, 2018, 38(10):3833-3843.
	Wei Xiaoxiao, Wang Xiaoming, Li Lei, et al. Temporal and spatial characteristics of municipal solid waste generation and treatment in China from 1979 to 2015[J]. China Environmental Science, 2018, 38(10):3833-3843.
[2]	梁建胜, 温贺平. 基于深度学习的视频关键帧提取与视频检索[J]. 控制工程, 2019, 26(5):965-970.
	Liang Jiansheng, Wen Heping. Key frame abstraction and retrieval of videos based on deep learning[J]. Control Engineering of China, 2019, 26(5):965-970.
[3]	赵亚男, 吴黎明, 陈琦. 基于多尺度融合SSD 的小目标检测算法[J]. 计算机工程, 2020, 46(1):247-254.
	Zhao Yanan, Wu Liming, Chen Qi. Small object detection algorithm based on multi-scale fusion SSD[J]. Computer Engineering, 2020, 46(1):247-254.
[4]	Ma Y, Zhang P, Tang Y. Research on fish image classification based on transfer learning and convolutional neural network model[C]. Huangshan: Proceedings of the Fourteenth International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery, 2018.
[5]	张亚须, 龙晖, 云利军. 基于改进DPM模型的行人检测方法研究[J]. 大理大学学报, 2018, 3(6):13-18.
	Zhang Yaxu, Long Hui, Yun Lijun. Research on pedestrian detection method based on improved DPM model[J]. Journal of Dali University, 2018, 3(6):13-18.
[6]	Ren S, He K, Girshick R, et al. Faster R-CNN: Towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2017, 39(6):1137-1149.
[7]	闫新庆, 杨喻涵, 陆桂明. 基于改进Faster-RCNN的目标检测算法研究[J]. 中国新通信, 2021, 23(1):46-48.
	Yan Xinqing, Yang Yuhan, Lu Guiming. Research on target detection algorithm based on improved Fast-RCNN[J]. China New Communications, 2021, 23(1):46-48.
[8]	Wang H B, Zhang Z D. Text detection algorithm based on improved YOLOv3[C]. Beijing: Proceedings of IEEE the Ninth International Conference on Electronics Information and Emergency Communication, 2019.
[9]	Rezatofighi H, Tsoi N, Gwak J Y, et al. Generalized intersection over union: A metric and a loss for bounding box regression[C]. Long Beach: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019.
[10]	Gu S, Ding L. A complex-valued VGG network based deep learing algorithm for image recognition[C]. Wanzhou: Proceedings of the Ninth International Conference on Intelligent Control and Information Processing, 2018.
[11]	钟志权, 袁进, 唐晓颖. 基于卷积神经网络的左右眼识别[J]. 计算机研究与发展, 2018, 55(8):1667-1673.
	Zhong Zhiquan, Yuan Jin, Tang Xiaoying. Left-vs-Right eye discrimination based on convolutional neural network[J]. Journal of Computer Research and Development, 2018, 55(8):1667-1673.
[12]	柯岩, 林小竹, 廖蕊, 等. 卷积神经网络层级分解研究[J]. 计算机工程, 2019, 45(11):191-197.
	Ke Yan, Lin Xiaozhu, Liao Rui, et al. Research on hierarchical decomposition of convolutional neural network[J]. Computer Engineering, 2019, 45(11):191-197.
[13]	Cao Z, Long M, Wang J, et al. HashNet: Deep learning to hash by continuation[C]. Venice: Proceedings of the IEEE International Conference on Computer Vision, 2017.
[14]	王一宁, 秦品乐, 李传朋, 等. 基于残差神经网络的图像超分辨率改进算法[J]. 计算机应用, 2018, 38(1):246-254. doi: 10.11772/j.issn.1001-9081.2017061461
	Wang Yining, Qin Pinle, Li Chuanpeng. Improved algorithm of image super resolution based on residual neural network[J]. Journal of Computer Applications, 2018, 22(1):246-254. doi: 10.11772/j.issn.1001-9081.2017061461
[15]	汪欣, 吴薇, 曾照. 基于视频的人脸检测算法研究[J]. 电子科技, 2020, 33(2):25-31.
	Wang Xin, Wu Wei, Zeng Zhao. Research on face detection algorithm based on video[J]. Electronic Science and Technology, 2020, 33(2):25-31.
[16]	秦莹华, 李菲菲, 陈虬. 基于迁移学习的多标签图像标注[J]. 电子科技, 2018, 31(8):21-24.
	Qin Yinghua, Li Feifei, Chen Qiu. Multi-label image annotation based on transfer learning[J]. Electronic Science and Technology, 2018, 31(8):21-24.
[17]	Weng Q, Mao Z, Lin J, et al. Land-use classification via extreme learning classifier based on deep convolutional features[J]. IEEE Geoscience and Remote Sensing Letters, 2017, 14(5):704-708. doi: 10.1109/LGRS.2017.2672643
[18]	Rauf H T, Lali M I U, Zahoor S, et al. Visual features based automated identification of fish species using deep convolutional neural networks[J]. Computers and Electronics in Agriculture, 2019, 16(7):154-167.

垃圾种类	YOLOv4	Proposed model
Cans	90.21%	94.56%
Cigarette	89.62%	90.85%
CigaretteCase	89.67%	89.12%
FruitPeel	83.52%	86.27%
Package	91.37%	93.81%
Paper	81.64%	85.24%
PlasticBottle	87.41%	93.73%
SeedSheel	76.52%	77.39%
Leaf	84.63%	87.24%

backbone	卷积层	感受野	参数
CSPResNEXt50	16	425×425	20.6 MB
CSPDarknet53	29	725×725	27.6 MB

网络模型	K-means	Attention	map@0.5
YOLOv4	×	×	85.26%
YOLOv4-1	√	×	86.34%
Proposed model	√	√	88.07%