面向复杂背景的CFFA—ResNet苹果叶片病害识别模型构建

doi:10.13733/j.jcam.issn.2095-5553.2025.06.025

摘要/Abstract

摘要：

目前大部分叶片病害识别研究在简单背景下的公开数据集中进行，实际应用过程中叶片背景复杂且数据样本很少，传统网络模型难以从复杂背景图像中有效提取病害区域特征。基于经典ResNet模型，提出一种全新的跨层特征融合注意力网络CFFA—ResNet。通过双分支跨层连接提取并融合不同维度特征，实现上下文信息的传递，增强细微判别性特征间的关联；通过注意力特征融合实现局部与全局语义信息的互补，并以加权的方式突出重要信息对融合的影响，重点提取病害区域特征，降低无关信息的干扰。结果表明，相同的试验环境下，与常见传统模型相比，新模型平均识别准确率可达97.45%，分类性能明显提升，说明该网络有助于复杂背景下叶片病害识别的研究。

关键词: 智慧农业, 苹果叶片病害, 残差网络, 特征融合, 注意力机制, 复杂背景

Abstract:

Current research on leaf disease identification is predominantly carried out using public data sets with simplified contexts. However, in real‑world practical applications, the leaf backgrounds are often complex, and data samples are limited. Traditional network models struggle to effectively extract disease‑specific features from images with such challenging backgrounds. To address this issue, a novel cross‑layer feature fusion attention network (CFFA—ResNet) is proposed, building upon the classic ResNet architecture. The CFFA—ResNet model extracted and fused features of different dimensions through dual‑branch and cross‑layer connections to realize the transmission of contextual information and enhance the relationship between subtle discriminative features. By fusing attention features, the model combined local and global semantic information, thereby highlighting the contribution of important information through weighted fusion while focusing on the extraction of disease‑specific features while mitigating interference from irrelevant data. The experimental results show that, under identical experimental conditions, the CFFA—ResNet achieved an average recognition accuracy of 97.45%, outperforming traditional models and the significantly enhancing classification performance. These findings indicate that CFFA—ResNet is a robust and effective tool for leaf disease identification in complex contexts.

Key words: smart agriculture, apple leaf disease, residual network, feature fusion, attention mechanism, complex background

中图分类号:

裴文杰, 刘拥民, 胡魁, 石婷婷, . 面向复杂背景的CFFA—ResNet苹果叶片病害识别模型构建[J]. 中国农机化学报, 2025, 46(6): 169-175.

Pei Wenjie, , Liu Yongmin, , Hu Kui, , Shi Tingting, . Construction of CFFA—ResNet apple leaf disease identification model for complex background[J]. Journal of Chinese Agricultural Mechanization, 2025, 46(6): 169-175.

参考文献

[ 1 ] 王璇, 刘军弟, 邵砾群, 等. 我国苹果产业年度发展状况及其趋势与建议[J]. 中国果树, 2018(3): 101-104, 108.

[ 2 ] 叶中华, 赵明霞, 贾璐. 复杂背景农作物病害图像识别研究[J]. 农业机械学报, 2021, 52(S1): 118-124, 147.

Ye Zhonghua, Zhao Mingxia, Jia Lu. Image recognition of crop diseases in complex background [J]. Transactions of the Chinese Society for Agricultural Machinery, 2021, 52(S1): 118-124, 147.

[ 3 ] 苏仕芳, 乔焰. 基于迁移学习的葡萄叶片病害识别及移动端应用[J]. 农业工程学报, 2021, 37(10): 127-134.

Su Shifang, Qiao Yan. Recognition of grape leaf diseases and mobile application based on transfer learning [J]. Transactions of the Chinese Society of Agricultural Engineering, 2021, 37(10): 127-134.

[ 4 ] 彭红星, 徐慧明, 刘华鼐. 融合双分支特征和注意力机制的葡萄病虫害识别模型[J]. 农业工程学报, 2022, 38(10): 156-165.

Peng Hongxing, Xu Huiming, Liu Huanai. Model for identifying grape pests and diseases based on two‑branch feature fusion and attention mechanism [J]. Transactions of the Chinese Society of Agricultural Engineering, 2022, 38(10): 156-165.

[ 5 ] 苏宝峰, 沈磊, 陈山, 等. 基于注意力机制的葡萄品种多特征分类方法[J]. 农业机械学报, 2021, 52(11): 226-233, 252.

Su Baofeng, Shen Lei, Chen Shan, et al. Multi‑features identification of grape cultivars based on attention mechanism [J]. Transactions of the Chinese Society for Agricultural Machinery, 2021, 52(11): 226-233, 252.

[ 6 ] 胡志伟, 杨华, 黄济民, 等. 基于注意力残差机制的细粒度番茄病害识别[J]. 华南农业大学学报, 2019, 40(6): 124-132.

Hu Zhiwei, Yang Hua, Huang Jimin, et al. Fine‑grained tomato disease recognition based on attention residual mechanism [J]. Journal of South China Agricultural University, 2019, 40(6): 124-132.

[ 7 ] 张文轩, 吴秦. 基于多分支注意力增强的细粒度图像分类[J]. 计算机科学, 2022, 49(5): 105-112.

[ 8 ] 宋丽娟. 基于图像的农作物病害识别关键算法研究[D]. 西安: 西北大学, 2018.

[ 9 ] 罗建豪, 吴建鑫. 基于深度卷积特征的细粒度图像分类研究综述[J]. 自动化学报, 2017, 43(8): 1306-1318.

[10] 龙辰志, 陈平, 李传坤. 融合全局—局部上下文信息的小目标多人姿态估计[J]. 计算机工程, 2024, 50(4): 342-349.

[11] 齐爱玲, 王宣淋. 基于中层细微特征提取与多尺度特征融合细粒度图像识别[J]. 计算机应用, 2023, 43(8): 2556-2563.

[12] He K, Zhang X, Ren S, et al. Deep residual learning for image recognition [C]. IEEE Conference on Computer Vision and Pattern Recognition, 2016: 770-778.

[13] 黄英来, 艾昕. 改进残差网络在玉米叶片病害图像的分类研究[J]. 计算机工程与应用, 2021, 57(23): 178-184.

Huang Yinglai, Ai Xin. Research on classification of corn leaf disease image by improved residual network [J]. Computer Engineering and Applications, 2021, 57(23): 178-184.

[14] Zagoruyko S, Komodakis N. Wide residual networks [J]. arVix preprint arVix: 1605.07146, 2016.

[15] Tan M, Pang R, Le Q V. EfficientDet: Scalable and efficient object detection [C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2020: 10778-10787.

[16] Woo S, Park J, Lee J, et al. CBAM: Convolutional block attention module [C]. European Conference on Computer Vision (ECCV), 2018: 1-17.

[17] Li F, Deng J, Li K. ImageNet: Constructing a large‑scale image database [J]. Journal of Vision, 2010, 9(8): 1037.

[18] Zhai S, Wu H, Kumar A, et al. S3Pool: Pooling with stochastic spatial sampling [C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017: 4003-4011.

[1]	王江晴, 马春, 莫海芳, 帖军, 田娟娟, . 基于融合双注意力机制的野生菌图像识别方法[J]. 中国农机化学报, 2025, 46(6): 70-76.
[2]	邹小波, 高文健, 石吉勇, 史永强, 申婷婷. 基于改进DenseNet与迁移学习的食品图像分类技术[J]. 中国农机化学报, 2025, 46(6): 77-84.
[3]	邵明明, 张立萍, 郑威强, 吕文涛, 雷佳乐. 基于改进YOLOv7的复杂环境下香梨和果梗的采摘识别[J]. 中国农机化学报, 2025, 46(6): 128-135.
[4]	崔岩, 庄卫东, 秦韬, 王楠. 基于改进YOLOv5的玉米植株检测与识别研究[J]. 中国农机化学报, 2025, 46(6): 136-141.
[5]	甄又陈, 王佳宇, 王宁, 刘升平, 林克剑, 李艳艳. 基于改进ResNet50的草原蝗虫种类智能识别APP系统#br#[J]. 中国农机化学报, 2025, 46(5): 68-78.
[6]	刘坤, , 吉宏亚, 黄程菲, 王晓, 朱一帆. 基于改进YOLOv5s的番茄成熟度识别技术研究[J]. 中国农机化学报, 2025, 46(5): 79-85.
[7]	王靖, 姜文刚, 程耀, 钱伟. 基于Lab颜色空间改进U—Net的稻田杂草分割方法[J]. 中国农机化学报, 2025, 46(5): 148-154.
[8]	陈佳豪, , 汪语哲, , 段晓东, , 梁凯华, . 基于改进YOLOv5s的青稞病虫害检测方法[J]. 中国农机化学报, 2025, 46(5): 162-168.
[9]	陈浩宇, 苗玉彬. 基于GASF变换和深度学习的柑橘内部品质分析[J]. 中国农机化学报, 2025, 46(4): 133-138.
[10]	赵玉清, 焦雨杰, 李宏, 王天允, 李嘉舜, 张悦, . 改进ShuffleNet V1算法的咖啡豆分级方法研究[J]. 中国农机化学报, 2025, 46(4): 194-203.
[11]	吕宗旺, 王甜甜, 孙福艳, 祝玉华, . 基于改进YOLOv8m的小麦仓储粮虫检测方法[J]. 中国农机化学报, 2025, 46(3): 108-114.
[12]	张俊宁, 闫英, 胡欢, 毕泽洋, 邢宇. 基于Sim—YOLO的复杂背景下樱桃番茄多目标检测[J]. 中国农机化学报, 2025, 46(3): 238-245.
[13]	郭敬涛, 吕凤, 章慧婷, 杨彪, 刘大洋. 基于迁移学习与轻量化YOLOv5s的草莓目标检测方法[J]. 中国农机化学报, 2025, 46(3): 253-260.
[14]	颜士军, 朱红梅, 王雅童, 张亮. 基于字词融合和注意力机制的兽药文本命名实体识别#br#[J]. 中国农机化学报, 2025, 46(3): 336-342.
[15]	谷瑞, 顾家乐, 宋翠玲, 钱春花. 基于坐标注意力的多尺度轻量级苹果叶片病害识别模[J]. 中国农机化学报, 2025, 46(2): 173-180.