|本期目录/Table of Contents|

[1]杨德龙,李婧.基于注意力与小平方核的ConvNeXt农业杂草识别方法[J].江苏农业科学,2024,52(14):207-214.
 Yang Delong,et al.ConvNeXt agricultural weed recognition method based on attention and small square kernel[J].Jiangsu Agricultural Sciences,2024,52(14):207-214.
点击复制

基于注意力与小平方核的ConvNeXt农业杂草识别方法(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14期
页码:
207-214
栏目:
农业工程与信息技术
出版日期:
2024-07-20

文章信息/Info

Title:
ConvNeXt agricultural weed recognition method based on attention and small square kernel
作者:
杨德龙李婧
上海电力大学计算机科学与技术学院,上海 200120
Author(s):
Yang Delonget al
关键词:
杂草识别ConvNeXtGRN正则化策略小平方核通道注意力
Keywords:
-
分类号:
S126;TP391.41
DOI:
-
文献标志码:
A
摘要:
针对复杂自然环境下杂草识别准确率不高、泛化和拟合能力较差等问题,提出基于注意力与小平核的ConvNeXt杂草图像识别方法。首先,在ConvNeXt模块中加入GRN正则化策略,有效减少识别过程的过拟合风险,提高模型的泛化能力;其次,提出在所有Block中把7×7的深度可分离卷积分解为有4个平行分支的小平方核,提升对杂草图像的特征提取能力;最后,在ConvNeXt结合上述方法下,引入SENet通道注意力模块,进一步提高模型在通道方向的特征融合,强化杂草特征,构建出杂草识别模型。为验证模型的识别性能,使用公开的9类杂草图像样本进行对比试验,结果表明,与主流模型相比,模型在准确率、精确率、召回率、F1分数上均表现优异,分别达到96.172%、95556%、96.478%、96.014%;消融试验结果表明,与基准模型ConvNeXt相比,GRN、小平方核分别提高8.639%、5691%,SENet在前二者基础上提高了5.174百分点;可视化分析证明,引入的通道注意力能更好关注到杂草特征。该模型可提高杂草识别准确率和对真实环境的泛化能力,为精准防控杂草提供有效的解决方法。
Abstract:
-

参考文献/References:

[1]Tauber M,Gollan B,Schmittner C,et al. Passive precision farming reshapes the agricultural sector[J]. Computer,2023,56(1):120-124.
[2]李东升,胡文泽,兰玉彬,等. 深度学习在杂草识别领域的研究现状与展望[J]. 中国农机化学报,2022,43(9):137-144.
[3]赵辉,曹宇航,岳有军,等. 基于改进DenseNet的田间杂草识别[J]. 农业工程学报,2021,37(18):136-142.
[4]疏雅丽,张国伟,王博,等. 基于深层连接注意力机制的田间杂草识别方法[J]. 计算机工程与应用,2022,58(6):271-277.
[5]曲福恒,李婉婷,杨勇,等. 基于图像增强和注意力机制的作物杂草识别[J]. 计算机工程与设计,2023,44(3):815-821.
[6]Olsen A,Konovalov D A,Philippa B,et al. DeepWeeds:a multiclass weed species image dataset for deep learning[J]. Scientific Reports,2019,9:2058.
[7]Liu Z,Mao H Z,Wu C Y,et al. A ConvNet for the 2020s[C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New Orleans:IEEE,2022:11966-11976.
[8]Woo S,Debnath S,Hu R H,et al. ConvNeXt v2:co-designing and scaling ConvNets with masked autoencoders[C]//2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Vancouver:IEEE,2023:16133-16142.
[9]Hu J,Shen L,Sun G. Squeeze-and-excitation networks[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Salt Lake City:IEEE,2018:7132-7141.
[10]Yu W H,Zhou P,Yan S C,et al. InceptionNeXt:when inception meets ConvNeXt[EB/OL]. [2022-11-12]. http://arxiv.org/abs/2303.16900.
[11]Alom M Z,Taha T M,Yakopcic C,et al. The history began from AlexNet:a comprehensive survey on deep learning approaches[EB/OL]. [2022-11-12]. http://arxiv.org/abs/1803.01164.
[12]Szegedy C,Vanhoucke V,Ioffe S,et al. Rethinking the inception architecture for computer vision[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas:IEEE,2016:2818-2826.
[13]Yu W,Yang K,Bai Y,et al. Visualizing and comparing AlexNet and VGG using deconvolutional layers[C]//Proceedings of the 33rd International Conference on Machine Learning,2016.
[14]Xia X L,Xu C,Nan B. Inception-v3 for flower classification[C]//2017 2nd International Conference on Image,Vision and Computing. Chengdu:IEEE,2017:783-787.
[15]Ma N N,Zhang X Y,Zheng H T,et al. ShuffleNet v2:practical guidelines for efficient CNN architecture design[C]//Computer Vision -ECCV 2018 of Munich:ACM,2018:122-138.
[16]Loshchilov I,Hutter F. Decoupled weight decay regularization[EB/OL]. [2022-11-12]. http://arxiv.org/abs/1711.05101.
[17]You Y,Gitman I,Ginsburg B. Large batch training of convolutional networks[EB/OL]. [2022-11-12]. http://arxiv.org/abs/1708.03888.
[18]李文举,苏攀,崔柳. 基于随机扰动的过拟合抑制算法[J]. 计算机仿真,2022,39(5):134-138.
[19]Ho Y,Wookey S.The real-world-weight cross-entropy loss function:modeling the costs of mislabeling[J]. IEEE Access,2020,8:4806-4813.
[20]王改华,翟乾宇,曹清程,等. 基于MoblieNet v2的图像语义分割网络[J]. 陕西科技大学学报,2022,40(1):174-181.
[21]朱炳宇,刘朕,张景祥.融合Grad-CAM和卷积神经网络的COVID-19检测算法[J]. 计算机科学与探索,2022,16(9):2108-2120.

相似文献/References:

[1]胡盈盈,王瑞燕,郭鹏涛,等.基于近地光谱特征的玉米田间杂草识别研究[J].江苏农业科学,2020,48(8):242.
 Hu Yingying,et al.Recognition of weeds in maize fields based on near-earth spectrum characteristics[J].Jiangsu Agricultural Sciences,2020,48(14):242.
[2]李彧,余心杰,郭俊先.基于全卷积神经网络方法的玉米田间杂草识别[J].江苏农业科学,2022,50(6):93.
 Li Yu,et al.Weed recognition in corn field based on fully convolutional neural network (FCN) method[J].Jiangsu Agricultural Sciences,2022,50(14):93.
[3]曾晏林,贺壹婷,蔺瑶,等.基于BCE-YOLOv5的苹果叶部病害检测方法[J].江苏农业科学,2023,51(15):155.
 Zeng Yanlin,et al.An apple leaf disease detection method based on BCE-YOLOv5[J].Jiangsu Agricultural Sciences,2023,51(14):155.
[4]马晓,邢雪,武青海.基于改进ConvNext的复杂背景下玉米叶片病害分类[J].江苏农业科学,2023,51(19):190.
 Ma Xiao,et al.Maize leaf disease classification under complex background based on improved ConvNext[J].Jiangsu Agricultural Sciences,2023,51(14):190.
[5]李豫晋,沈陆明,何少芳,等.基于改进MobileNet v3的苹果叶片病害识别研究[J].江苏农业科学,2024,52(12):224.
 Li Yujin,et al.Identification of apple leaf diseases based on improved MobileNet v3[J].Jiangsu Agricultural Sciences,2024,52(14):224.
[6]高发瑞,古华宁,张巧玲,等.基于农业大数据和深度学习的稻田杂草识别[J].江苏农业科学,2024,52(18):215.
 Gao Farui,et al.Identification of weeds in rice field based on agricultural big data and deep learning[J].Jiangsu Agricultural Sciences,2024,52(14):215.
[7]黄友锐,王小桥,韩涛,等.基于改进YOLO v8n的甜菜杂草检测算法研究[J].江苏农业科学,2024,52(24):196.
 Huang Yourui,et al.A detection method for sugar beets and weeds based on improved YOLO v8n algorithm[J].Jiangsu Agricultural Sciences,2024,52(14):196.

备注/Memo

备注/Memo:
收稿日期:2023-08-17
基金项目:国家自然科学基金(编号:U1936123)。
作者简介:杨德龙(1999—),男,河南信阳人,硕士研究生,主要从事深度学习、图像识别研究。E-mail:delongyang@mail.shiep.edu.cn。
通信作者:李婧,博士,副教授,主要从事深度学习、区块链技术研究。E-mail:lijing@shiep.edu.cn。
更新日期/Last Update: 2024-07-20