[1]王忠培,谢成军,董伟,等.基于多维间注意力机制的水稻病害识别模型[J].江苏农业学报,2024,(04):625-635.[doi:doi:10.3969/j.issn.1000-4440.2024.04.006]
 WANG Zhong-pei,XIE Cheng-jun,DONG Wei,et al.Rice disease identification model based on multi-dimensional attention mechanism[J].,2024,(04):625-635.[doi:doi:10.3969/j.issn.1000-4440.2024.04.006]
点击复制

基于多维间注意力机制的水稻病害识别模型()
分享到:

江苏农业学报[ISSN:1006-6977/CN:61-1281/TN]

卷:
期数:
2024年04期
页码:
625-635
栏目:
植物保护
出版日期:
2024-04-30

文章信息/Info

Title:
Rice disease identification model based on multi-dimensional attention mechanism
作者:
王忠培1 谢成军2 董伟1 管博伦1
(1.安徽省农业科学院农业经济与信息研究所,安徽合肥230001;2.中国科学院合肥智能机械研究所,安徽合肥230031)
Author(s):
WANG Zhong-pei1XIE Cheng-jun2DONG Wei1GUAN Bo-lun1
(1.Institute of Agricultural Economics and Information, Anhui Academy of Agricultural Sciences, Hefei 230001, China;2.Hefei Institute of Intelligent Machinery, Chinese Academy of Sciences, Hefei 230031, China)
关键词:
水稻病害三维注意力多维间关系注意力机制识别
Keywords:
rice diseasesthree-dimensional attentionmulti-dimensional relationshipattention mechanismidentification
分类号:
TP391.41
DOI:
doi:10.3969/j.issn.1000-4440.2024.04.006
摘要:
水稻病害的快速、准确识别是水稻病害防治的前提,也是提高水稻产量和品质的有效途径之一。为了提高水稻病害识别的准确率,本研究提出一种多维间的三维注意力水稻病害识别模型Inter_3DRiceNet网络模型,通过3个不同维度(通道维度、高度维度以及宽度维度)提取水稻病害特征信息。通道维度主要构建基于通道关系的三维立体注意力机制,通过建立一维的通道间关系注意力机制再结合二维空间关系,最终获得基于通道关系的三维注意力特征信息。高度维度建立的是基于高度维度关系的三维注意力机制,而宽度维度建立的是基于宽度维度关系的立体注意力机制。然后将以上3个不同维度的注意力信息进行简单的相加再取平均值作为最终的病害提取特征。通过这种方式,不仅可以获取输入图像更丰富的特征,而且可以获得不同维度的立体空间关系。试验结果表明,在自建的6种真实自然环境水稻病害数据集中,本研究提出的Inter_3DRiceNet网络模型在测试集取得了98.32%的最高准确率,高于经典网络模型ResNet34、ResNet50、MobileNetV2、DenseNet、EfficientNet_B0和通道注意力机制模型SENet和GCT。可见本研究方法有效提高了水稻病害的识别准确率,获得了优于经典网络模型和通道注意力模型的识别准确率,有助于提升自然环境下对常见水稻病害的识别性能。
Abstract:
Rapid and accurate identification of rice diseases is a prerequisite for controlling rice diseases and is one of the effective ways to improve rice yield and quality. To improve the identification accuracy of rice diseases, a network model of multi-dimensional attention mechanism for rice disease identification named Inter_3DRiceNet was proposed in this study to extract rice disease feature information through three different dimensions (channel dimension, height dimension and width dimension). The channel dimension mainly constructed a three-dimensional attention mechanism based on channel relationship, and finally obtained three-dimensional attention feature information based on channel relationship by establishing a one-dimensional attention mechanism of inter-channel relationship combined with two-dimensional spatial relationship. The height dimension established a three-dimensional attention mechanism based on the height dimension relationship, while the width dimension established a tridimensional attention mechanism based on the width dimension relationship. The attention information of the above three different dimensions was simply summed and then averaged as the final disease extraction features. Thus, besides more abundant features of the input images could be obtained, stereoscopic spatial relations of different dimensions could also be obtained. The experimental results showed that, the Inter_3DRiceNet network model proposed in the study got the highest accuracy of 98.32% in the test sets of the six self-constructed rice disease datasets in real natural environment, which was higher than the classical network models such as ResNet34, ResNet50, MobileNetV2, DenseNet, EfficientNet_B0, and channel attention mechanism models SENet and GCT. The research method improved the recognition accuracy of rice diseases effectively and obtained better classification accuracy than the classical network model and the channel attention model, which can help improve the performance of common rice diseases recognition in natural environment.

参考文献/References:

[1]佚名. 2022年全国农作物重大病虫害发生趋势预报[J]. 中国植保导刊,2022,42(4):107-108.
[2]赵立新,侯发东,吕正超,等. 基于迁移学习的棉花叶部病虫害图像识别[J]. 农业工程学报,2020,36(7):184-191.
[3]刘阳,高国琴. 采用改进的 SqueezeNet 模型识别多类叶片病害[J]. 农业工程学报,2021,37(2):187-195.
[4]鲍文霞,吴德钊,胡根生,等. 基于轻量型残差网络的自然场景水稻害虫识别[J]. 农业工程学报,2021,37(16):145-152.
[5]ESPEJO-GARCIA B, MALOUNAS I, MYLONAS N, et al. Using EfficientNet and transfer learning for image-based diagnosis of nutrient deficiencies[J]. Computers and Electronics in Agriculture,2022,196:106868.
[6]HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]. Piscataway: IEEE,2018:7132-7141.
[7]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.
[8]WANG Q, WU B, ZHU P, et al. ECA-Net: efficient channel attention for deep convolutional neural networks[C]. Piscataway:IEEE,2020:11531-11539.
[9]孙俊,朱伟栋,罗元秋,等. 基于改进 MobileNet-V2 的田间农作物叶片病害识别[J]. 农业工程学报,2021,37(22):161-169.
[10]甘雨,郭庆文,王春桃,等. 基于改进 EfficientNet 模型的作物害虫识别[J]. 农业工程学报,2022,38(1):203-211.
[11]HOU Q, ZHOU D, FENG J. Coordinate attention for efficient mobile network design[C]. Piscataway:IEEE, 2021.
[12]WU X, ZHAN C, LAI Y K, et al. IP102: a large-scale benchmark dataset for insect pest recognition[C]. Piscataway:IEEE, 2019.
[13]ZHAO Y, SUN C, XU X, et al. RIC-Net: a plant disease classification model based on the fusion of Inception and residual structure and embedded attention mechanism[J]. Computers and Electronics in Agriculture, 2022,193:106644.
[14]ZHAO X, LI K, LI Y, et al. Identification method of vegetable diseases based on transfer learning and attention mechanism[J]. Computers and Electronics in Agriculture,2022,193:106703.
[15]HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition[C]. Piscataway:IEEE,2016:770-778.
[16]SANDLER M, HOWARD A, ZHU M, et al. Mobilenetv2: inverted residuals and linear bottlenecks[C]. Piscataway: IEEE,2018:4510-4520.
[17]HUANG G, LIU Z, VAN DER MAATEN L, et al. Densely connected convolutional networks[C]. Piscataway:IEEE,2017.
[18]TAN M, LE Q. Efficientnet: rethinking model scaling for convolutional neural networks[C]. Madison:ACM,2019.
[19]YANG Z, ZHU L, WU Y, et al. Gated channel transformation for visual recognition[C]. Piscataway:IEEE,2020.
[20]SELVARAJU R R, COGSWELL M, DAS A, et al. Grad-cam: visual explanations from deep networks via gradient-based localization[C]. Piscataway:IEEE,2017.

相似文献/References:

[1]陆煜,俞经虎,朱行飞,等.基于卷积神经网络的轻量级水稻叶片病害识别模型[J].江苏农业学报,2024,(02):312.[doi:doi:10.3969/j.issn.1000-4440.2024.02.013]
 LU Yu,YU Jing-hu,ZHU Xing-fei,et al.A lightweight rice leaf disease recognition model based on convolutional neural network[J].,2024,(04):312.[doi:doi:10.3969/j.issn.1000-4440.2024.02.013]

备注/Memo

备注/Memo:
收稿日期:2023-02-24基金项目:国家自然科学基金项目(32171888)作者简介:王忠培(1981-),男,安徽金寨人,博士,助理研究员,研究方向为智能农业技术。(E-mail)wangzhongpei@aaas.org.cn通讯作者:谢成军, (E-mail)cjxie@iim.ac.cn
更新日期/Last Update: 2024-05-22