|本期目录/Table of Contents|

[1]蔡易南,肖小玲.基于改进YOLO v5n的葡萄叶病虫害检测模型轻量化方法[J].江苏农业科学,2024,52(7):198-205.
 Cai Yinan,et al.Lightweight method of grape leaf diseases and insect pests detection model based on improved YOLO v5n[J].Jiangsu Agricultural Sciences,2024,52(7):198-205.
点击复制

基于改进YOLO v5n的葡萄叶病虫害检测模型轻量化方法(PDF)
分享到:

《江苏农业科学》[ISSN:1002-1302/CN:32-1214/S]

卷:
第52卷
期数:
2024年第7期
页码:
198-205
栏目:
农业工程与信息技术
出版日期:
2024-04-05

文章信息/Info

Title:
Lightweight method of grape leaf diseases and insect pests detection model based on improved YOLO v5n
作者:
蔡易南肖小玲
长江大学计算机科学学院,湖北荆州 434000
Author(s):
Cai Yinanet al
关键词:
葡萄叶病害YOLO v5Slimming剪枝WIoU损失函数CARAFE算子
Keywords:
-
分类号:
TP391.41;S126
DOI:
-
文献标志码:
A
摘要:
由于较大的参数量和较高的计算复杂度,直接在移动端部署通用检测及识别模型的难度较高。为了解决轻量化的移动端部署难题及提升移动设备上葡萄叶病害的检测能力,拟提出1种轻量化、高精度、实时性的检测模型。首先,引入Slimming算法对传统的YOLO v5n网络进行缩减,利用模型稀疏化训练、批归一化的缩放因子分布状况对不重要的通道进行筛选;其次,引入轻量级上采样算子CARAFE增加感受野,进行数据特征融合;最后,将边界框回归损失函数改进为WIoU损失函数,制定合适的梯度增益分配策略来获得更加精准的框定位提升模型对每个类别目标的检测能力。试验结果表明,改进后的模型能够在保持模型性能的情况下有效轻量化。与传统的YOLO v5n相比,改进后的算法mAP提高了0.2百分点,同时改进后的模型权重、参数量、计算量分别为1.6 MB、0.6 M、1.8 G,分别比原模型减少了58%、67%、57%,能够满足移动端和嵌入式设备的部署要求。
Abstract:
-

参考文献/References:

[1]田野,陈冠铭,李家芬,等. 世界葡萄产业发展现状[J]. 热带农业科学,2018,38(6):96-101.
[2]翟肇裕,曹益飞,徐焕良,等. 农作物病虫害识别关键技术研究综述[J]. 农业机械学报,2021,52(7):1-18.
[3]戴久竣,马肄恒,吴坚,等. 基于改进残差网络的葡萄叶片病害识别[J]. 江苏农业科学,2023,51(5):208-215.
[4]谢圣桥,宋健,汤修映,等. 基于迁移学习和残差网络的葡萄叶部病害识别[J]. 农机化研究,2023,45(8):18-23,28.
[5]何欣,李书琴,刘斌. 基于多尺度残差神经网络的葡萄叶片病害识别[J]. 计算机工程,2021,47(5):285-291.
[6]陈浪浪,张艳. 基于改进深度卷积神经网络的水稻病虫害识别[J]. 山东农业科学,2023,55(5):164-172.
[7]张伏,陈自均,鲍若飞,等. 基于改进型YOLO v4-LITE轻量级神经网络的密集圣女果识别[J]. 农业工程学报,2021,37(16):270-278.
[8]刘广,胡国玉,赵腾飞,等. 基于改进YOLO v3的葡萄叶部病虫害检测方法[J]. 微电子学与计算机,2023,40(2):110-119.
[9]Ren S Q,He K M,Girshick R,et al. Faster R-CNN:towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2017,39(6):1137-1149.
[10]Dai J F,Li Y,He K M,et al. R-FCN:object detection via region-based fully convolutional networks[J]. Advances in Neural Information Processing Systems,2016,29:379-387.
[11]Liu W,Anguelov D,Erhan D,et al. SSD:single shot multibox detector[C]//Computer Vision–ECCV 2016:14th European Conference,Amsterdam:Springer International Publishing,2016:21-37.
[12]Bochkovskiy A,Wang C Y,Liao H Y M. YOLO v4:optimal speed and accuracy of object detection[J/OL]. [2023-09-09]https://arxiv.org/abs/2004. 10934.
[13]Redmon J,Divvala S,Girshick R,et al. You only look once:unified,real-time object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,Las Vegas,USA,2016:779-788.
[14]Redmon J,Farhadi A. Yolo9000:better,faster,stronger[C]//Proceedings of the IEEE conference on Computer Vision and Pattern Recognition,Honolulu,USA,2017:7263-7271.
[15]Redmon J,Farhadi A. YOLO v3:An incremental improvement[J/OL]. [2023-09-09]https://arxiv.org/abs/1804. 02767.
[16]Lin T Y,Dollár P,Girshick R,et al. Feature pyramid networks for object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,Honolulu,USA,2017:2117-2125.
[17]Liu S,Qi L,Qin H,et al. Path aggregation network for instance segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,Salt Lake City,USA,2018:8759-8768.
[18]Liu Z,Li J,Shen Z,et al. Learning efficient convolutional networks through network slimming[C]//Proceedings of the IEEE International Conference on Computer Vision,Venice,Italy,2017:2736-2744.
[19]Tong Z,Chen Y,Xu Z,et al. Wise-iou:bounding box regression loss with dynamic focusing mechanism[J/OL]. [2023-09-09]https://arxiv.org/abs/2301. 10051.
[20]Wang J,Chen K,Xu R,et al. Carafe:content-aware reassembly of features[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision,Seoul,Korea(South),2019:3007-3016.
[21]Howard A,Sandler M,Chu G,et al. Searching for mobilenetv3[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision,Seoul,Korea(South),2019:1314-1324.
[22]Han K,Wang Y,Tian Q,et al. Ghostnet:more features from cheap operations[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,Seattle,USA,2020:1580-1589.
[23]Tan M,Le Q. Efficientnet:rethinking model scaling for convolutional neural networks[C]//International Conference on Machine Learning. PMLR,Long Beach,California,USA,2019:6105-6114.

相似文献/References:

[1]杜鹏程,蒋笃忠,向阳,等.基于YOLO v5s目标检测算法的烤烟鲜叶成熟度识别方法[J].江苏农业科学,2023,51(19):158.
 Du Pengcheng,et al.Identification method for fresh leaf maturity of flue-cured tobacco based on YOLO v5s target detection algorithm[J].Jiangsu Agricultural Sciences,2023,51(7):158.
[2]刘忠意,魏登峰,李萌,等.基于改进YOLO v5的橙子果实识别方法[J].江苏农业科学,2023,51(19):173.
 Liu Zhongyi,et al.Orange fruit recognition method based on improved YOLO v5[J].Jiangsu Agricultural Sciences,2023,51(7):173.
[3]李滨,樊健.基于YOLO v5的水稻害虫分类[J].江苏农业科学,2024,52(2):175.
 Li Bin,et al.Classification of rice pests based on YOLO v5[J].Jiangsu Agricultural Sciences,2024,52(7):175.
[4]温彬彬,张华,孟祥龙.基于改进YOLO v5的轻量化苹果检测方法[J].江苏农业科学,2024,52(12):217.
 Wen Binbin,et al.A lightweight apple detection method based on improved YOLO v5[J].Jiangsu Agricultural Sciences,2024,52(7):217.
[5]吴坚,秦玉广.基于改进YOLO v5的农田苗草检测方法[J].江苏农业科学,2024,52(13):197.
 Wu Jian,et al.Detection of seedlings and weeds in farmland based on an improved YOLO v5 algorithm[J].Jiangsu Agricultural Sciences,2024,52(7):197.
[6]李炳鑫,宋涛,高婕,等.基于YOLO v5模型的缺钙草莓叶片识别方法[J].江苏农业科学,2024,52(20):74.
 Li Bingxin,et al.Identification method of calcium-deficient strawberry leaves based on YOLO v5 model[J].Jiangsu Agricultural Sciences,2024,52(7):74.
[7]王博,胡蓉华.基于轻量化神经网络的桃树叶片病害检测方法[J].江苏农业科学,2025,53(5):138.
 Wang Bo,et al.A peach leaf disease detection method based on lightweight neural network[J].Jiangsu Agricultural Sciences,2025,53(7):138.

备注/Memo

备注/Memo:
收稿日期:2023-10-09
基金项目:国家自然科学基金(编号:61771354)。
作者简介:蔡易南(2000—),男,湖北黄冈人,硕士研究生,主要研究方向为计算机视觉与目标检测。E-mail:2022710621@yangtzeu.edu.cn。
通信作者:肖小玲,博士,教授,主要研究方向为智能信息处理与网络安全。E-mail:xxl@yangtzeu.edu.cn。
更新日期/Last Update: 2024-04-05