基于层间特征蒸馏网络的作物叶片病害检测

doi:10.13733/j.jcam.issn.2095-5553.2024.09.041

摘要/Abstract

摘要：

针对现有农作物叶片病害检测方法对有限标注样本利用不充分，导致模型识别精度不高、泛化性不强的问题，提出一种基于层间特征蒸馏网络的作物叶片病害检测方法。该方法采用支持分支和查询分支相互监督的元学习网络结构，首先，利用一组共享权重的特征提取网络将双分支网络的输入图片映射到深度特征空间，并采用多层下采样操作构造多尺度特征集；然后，在每层特征中计算自注意力机制，在层间计算交叉注意力机制，旨在强化不同尺度内和尺度间特征表达的鲁棒性和可靠性；最后，在跨尺度特征中引入知识蒸馏网络，旨在利用高层特征丰富浅层特征的语义信息，间接地增强不同尺度内和尺度间特征表达的鲁棒性。在马铃薯、苹果、番茄和玉米病害数据集上进行测试，所提方法分别获得 0. 953 1、0. 966 8、0. 955 2和 0. 954 2的识别精准率。

关键词: 病害叶片检测, 知识蒸馏, 交叉注意力, 自注意力, 知识反馈

Abstract:

Aiming at the problem of insufficient utilization of limited labeled samples in existing crop leaf disease detection methods，which leads to low recognition accuracy and weak generalizability of the model，a crop leaf disease detection method based on inter.layer feature distillation network is proposed. The method adopts a meta.learning network structure with a support branch and a query branch supervising each other. Firstly，a set of shared weight feature extraction networks are used to map the input images of the two branches to the deep feature space，and multi.scale feature sets are constructed by using multiple down.sampling operations. Then， self.attention mechanism is calculated in each layer feature， and cross.attention mechanism is calculated between layers，aiming to enhance the robustness and reliability of feature expression at different scales and between scales. Finally，a knowledge distillation network is introduced in the cross.scale features， aiming to enrich the semantic information of low.level features with high.level features indirectly，and further enhance the robustness of feature expression at different scales and between scales. The proposed method has achieved recognition accuracies of 0. 953 1，0. 966 8，0. 955 2 and 0. 954 2 on potato，apple，tomato and corn diseases，respectively.

Key words: disease leaf detection, knowledge distillation, cross.attention, self.attention, knowledge feedback ,

中图分类号:

S435
TP391

冯玉涵 , 孙剑 , 张志芳 . 基于层间特征蒸馏网络的作物叶片病害检测[J]. 中国农机化学报, 2024, 45(9): 271-277.

Feng Yuhan, Sun Jian, Zhang Zhifang. Crop leaf disease detection based on inter.layer feature distillation network [J]. Journal of Chinese Agricultural Mechanization, 2024, 45(9): 271-277.

参考文献

［ 1］杨佳昊，左昊轩，黄祺成，等 .基于 YOLOv5s的作物叶片病害检测模型轻量化方法［ J］.农业机械学报， 2023， 54（S1）：222-229.

Yang Jiahao， Zuo Haoxuan， Huang Qicheng， et al. Lightweight method for crop leaf disease detection model based on YOLOv5s［J］. Transactions of the Chinese Society forAgriculturalMachinery，2023，54（S1）：222-229.

［ 2］Huang X B，Chen A B，Zhou G X，et al. Tomato leaf disease detection system based on FC-SNDPN［J］. Multimedia ToolsandApplications，2023，82（2）：2121-2144.

［ 3］Kaur Prabhjot，Harnal Shilpi，Gautam Vinay，et al. A novel transfer deep learning method for detection and classification of plant leaf disease［J］. Journal of Ambient Intelligence and Humanized Computing，2023，14（9）： 12407-12424.

［ 4］李康顺，杨振盛，江梓锋，等 .基于改进 YOLOX-Nano的农作物叶片病害检测与识别方法［ J］.华南农业大学学报， 2023，44（4）：593-603.

Li Kangshun， Yang Zhensheng， Jiang Zifeng， et al. Detection and identification of crop leaf diseases based on improved YOLOX-Nano［J］. Journal of South China Agricultural University，2023，44（4）：593-603.

［ 5］刘敏，周丽 .基于多尺度特征融合网络的苹果病害叶片检测［ J］.中国农机化学报， 2023，44（8）：184-190.

Liu Min，Zhou Li. Apple disease leaf detection based on multi.scale feature fusion network［J］. Journal of Chinese Agricultural Mechanization，2023，44（8）：184-190.

［ 6］邢鹏康，李久朋 .基于小样本学习的马铃薯叶片病害检测［ J］.江苏农业科学， 2023，51（15）：203-210.

Xing Pengkang，Li Jiupeng. Potato leaf disease detection based on few.shot learning［J］. Jiangsu Agricultural Sciences，2023，51（15）：203-210.

［ 7］马丽，周巧黎，赵丽亚，等 .基于深度学习的番茄叶片病害分类识别研究［ J］.中国农机化学报， 2023，44（7）： 187-193，206.

Ma Li， Zhou Qiaoli， Zhao Liya， et al. Classification and recognition of tomato leaf diseases based on deep learning［J］. Journal of Chinese Agricultural Mechanization，2023，44（7）： 187-193，206.

［ 8］惠巧娟，孙婕 .基于多尺度特征度量元学习的玉米叶片病害识别模型研究［ J］.江苏农业科学， 2023，51（9）： 199-206.

Hui Qiaojuan，Sun Jie. Research on maize leaf disease recognition model based on multiscale feature metric meta-learning［J］. Jiangsu Agricultural Sciences，2023，51（9）：199-206.

［ 9］黄炜，王娟娟，殷学丽 .基于特征分离的小样本苹果病害叶片检测［ J］.江苏农业科学， 2023，51（23）：195-202. Huang Wei，Wang Juanjuan，Yin Xueli. Few.shot apple disease leaves detection based on feature extraction［J］. Jiangsu Agricultural Sciences，2023，51（23）：195-202.

［10］Bedi Punam，Gole Pushkar. Plant disease detection using hybrid model based on convolutional autoencoder and convolutional neural network［J］. Artificial Intelligence in Agriculture，2021，5：90-101.

［11］Zhao Yafeng，Chen Zhen，Gao Xuan，et al. Plant disease detection using generated leaves based on DoubleGAN［J］. IEEE/ACM Transactions on Computational Biology and Bioinformatics，2021，19（3）： 1817-1826.

［12］Ying Li，Sun Shiyu，Zhang Changshe，et al. One.stage disease detection method for maize leaf based on multi.scale feature fusion［J］. Applied Sciences，2022，12（16）： 7960.

［13］蒋清健，姚勇，王亚玲，等 .基于多尺度卷积神经网络算法的番茄叶片病害识别［ J］.江苏农业科学， 2023，51（15）：211-216.

Jiang Qingjiang，Yao Yong，Wang Yaling，et al. Tomato leaf disease recognition based on multi.scale convolutional neural network algorithm［J］. Jiangsu Agricultural Sciences，2023，51（15）：211-216.

［14］Tian Y N，Li E，Liang Z Z，et al. Diagnosis of typical apple diseases： A deep learning method based on multi.scale dense classification network［J］. Frontiers in Plant Science，2021，12：698474.

［15］马晓，董天亮，钟闻宇，等 .基于改进 ConvNeXt的大豆叶片病害分类研究［ J］.大豆科学， 2023，42（6）：733-741.

Ma Xiao， Dong Tianliang， Zhong Wenyu， et al. Soybean leaf diseases classification method based on improved ConvNeXt［J］. Soybean Science， 2023， 42（6）： 733-741.

［16］Lin H， Tse R， Tang S K， et al. Few-shot learning approach with multi.scale feature fusion and attention for plant disease recognition［J］. Frontiers in Plant Science， 2022，13：907916.

［17］李军，李志伟，李艳红 .基于多原型指导的小样本水稻害虫识别与分类［ J］.江苏农业科学， 2023，51（20）： 193-200.

Li Jun， Li Zhiwei， Li Yanhong. Few-shot rice pest recognition and classification based on multi.prototype guidance［J］. Jiangsu Agricultural Sciences， 2023， 51（20）：193-200.

［18］刘媛媛，王定坤，邬雷，等 .基于知识蒸馏和模型剪枝的轻量化模型植物病害识别［ J］.浙江农业学报， 2023，35（9）：2250-2264.

Liu Yuanyuan， Wang Dingkun， Wu Lei， et al. A light.weight model for plant disease identification based on model pruning and knowledge distillation［J］. Acta Agriculturae Zhejiangensis，2023，35（9）：2250-2264.

［19］Lin H，Tse R，Tang S K，et al. Few.shot learning for plant.disease recognition in the frequency domain［J］. Plants，2022，11（21）：2814.

［20］Sun J Q，Cao W，Fu X，et al. Few.shot learning for plant disease recognition： A review［J］. Agronomy Journal， 2023，11（2）：136-145.

[1]	李星逸, 付波, 范秀香, 权轶. 基于改进对抗蒸馏的轴承故障分类方法[J]. 中国农机化学报, 2024, 45(6): 178-183.
[2]	邢清桂, 吴凯, 周洪斌. 基于分层特征注意力解耦的农机轴承故障诊断[J]. 中国农机化学报, 2024, 45(5): 140-146.
[3]	陈启, 陈慈发, 邓向武, 袁单飞. 基于移动端轻量模型的杂草分类方法研究[J]. 中国农机化学报, 2022, 43(2): 163-170.