高光谱成像结合 BiTCN-SA的马铃薯晚疫病早期识别

Early Identification of Potato Late Blight Using Hyperspectral Imaging Combined with BiTCN-SA

摘要

及早识别晚疫病是控制马铃薯晚疫病发展的关键，为充分利用高光谱数据波段间特征信息，提高模型对马铃薯晚疫病早期识别的精度，本文提出一种基于双向时间卷积网络（BiTCN）融合自注意力机制（SA）的马铃薯晚疫病早期识别模型（BiTCN-SA）。BiTCN通过正向和反向卷积支路捕捉波段间相关性特征，充分利用前后波段的关联性；自注意力机制动态分配不同波段的重要性权重，提高关键波段对模型分类的贡献度。BiTCN-SA模型将自注意力与BiTCN相融合，在双向上实现局部卷积特征与全局注意力权重的结合，实现双重特征提取，提高模型识别精度。采集3个等级（健康、无症状期、症状初期）的叶片高光谱数据并建模分析，通过对比SVM、RF等机器学习方法和CNN、LSTM、TCN、BiTCN等深度学习方法，以验证本文模型优越性。结果表明，BiTCN-SA模型的收敛速度比单一TCN和BiTCN更快，且模型精度显著提高，比其他机器学习和深度学习方法，具备更强大的特征提取能力，总体准确率达到98%，且对无症状期的病叶识别率达到96%。该方法充分利用高光谱波段间的深层信息，且模型识别率相比于其他机器学习和深度学习方法有大幅提高，为马铃薯晚疫病早期预警和防治提供技术支持。

Abstract

Early identification of potato late blight is crucial for controlling its development. To fully utilize the inter-band characteristic information of hyperspectral data and improve the accuracy of models in the early identification of potato late blight, this study proposes a potato late blight early identification model (BiTCN-SA) based on a Bidirectional Temporal Convolutional Network (BiTCN) fused with a Self-Attention (SA) mechanism. The BiTCN captures inter-band correlation features through forward and backward convolution branches, and fully exploits the associations between preceding and subsequent bands. The self-attention mechanism dynamically assigns importance weights to different bands, enhancing the contribution of key bands to model classification. The BiTCN-SA model integrates self-attention with BiTCN to achieve a combination of local convolutional features and global attention weights in both directions, realizing dual feature extraction and improving the model's identification accuracy. This study collects hyperspectral potato leaf data from three stages (healthy, asymptomatic, and early symptomatic), conducts modeling and analysis. It verifies the superiority of the proposed model by comparing machine learning methods such as SVM and RF, and deep learning models including CNN, LSTM, TCN, and BiTCN. The results show that the BiTCN-SA model converges faster than standalone TCN and BiTCN models, with significantly improved accuracy. It demonstrates stronger feature extraction capability than other machine learning and deep learning methods, achieving an overall accuracy of 98% and an identification rate of 96% for asymptomatic diseased leaves. This method fully utilizes deep inter-band information from hyperspectral data, and its identification rate shows substantial improvement over other machine learning and deep learning methods, providing technical support for early warning and control of potato late blight.

关键词

马铃薯晚疫病 / 高光谱成像 / 早期识别 / 双向时间卷积网络 / 自注意力机制 / 特征提取

Key words

Potato late blight / hyperspectral imaging / early identification / bidirectional temporal convolutional network / self-attention mechanism / feature extraction

参考文献

[1] 侯冰茹, 刘鹏辉, 张洋, 等. 基于马铃薯叶片光纤光谱信息的晚疫病患病程度预测[J]. 光谱学与光谱分析, 2022, 42(05): 1426-1432.

[2] 罗孳孳, 武强, 朱玉涵, 等. 基于CARAH模型的重庆春马铃薯晚疫病侵染特征模拟分析[J]. 植物保护, 2024, 50(04): 80-90.

[3] 沈梦姣, 鲍浩, 张艳. 基于光谱-纹理特征的辣椒早疫病潜育期高光谱图像检测识别[J]. 生物化学与生物物理进展, 2025, 52(01): 233-243.

[4] Zhang N , Yang G J , Pan Y C , et al. A review of advanced technologies and sevelopment for hyperspectral-based plant disease detection in the past three decades[J]. Remote Sensing, 2020, 12(19): 34.

[5] Li X, Fu X L, Li H H . A CARS-SPA-GA feature wavelength selection method based on hyperspectral imaging with potato leaf disease classification[J]. Sensors, 2024, 24(20): 18.

[6] Tang Y, Yang J P, Zhuang J J, et al. Early detection of citrus anthracnose caused by Colletotrichum gloeosporioides using hyperspectral imaging[J]. Comput Electron Agric, 2023, 214: 11.

[7] Zhou Y X , Chen J Z , Ma J F , et al. Early warning and diagnostic visualization of Sclerotinia infected tomato based on hyperspectral imaging[J]. Scientific Reports, 2022, 12(1): 13.

[8] 潘健, 祁雁楠, 陈鲁威, 等. 基于可见/近红外高光谱成像技术的梨树叶部病害识别研究[J]. 中国农机化学报, 2024, 45(08): 162-169.

[9] Khan A, Vibhute A D, Mali S, et al. A systematic review on hyperspectral imaging technology with a machine and deep learning methodology for agricultural applications[J]. Ecological Informatics, 2022, 69: 12.

[10] Wang C Y , Liu B H , Liu L P , et al. A review of deep learning used in the hyperspectral image analysis for agriculture[J]. Artificial Intelligence Review, 2021, 54(7): 5205-5253.

[11] Chun S W, Song D J, Lee K H, et al. Deep learning algorithm development for early detection of Botrytis cinerea infected strawberry fruit using hyperspectral fluorescence imaging[J]. Postharvest Biology and Technology, 2024, 214: 15.

[12] 魏雷, 徐迎波, 曹海兵, 等. 基于AE-BiRNN 的片烟含水率高光谱检测[J]. 烟草科技, 2024, 57(02): 70-78.

[13] 邓昀, 牛照文, 冯琦尧, 等. 改进时间卷积网络的红壤有机质高光谱预测模型[J]. 光谱学与光谱分析, 2023, 43(09): 2942-2951.

[14] 吴叶兰, 管慧宁, 廉小亲, 等. 基于高光谱成像和Att-BiGRU-RNN 的柑橘病叶分类[J]. 农业机械学报, 2023, 54(01): 216-223.

[15] 徐胜勇, 刘政义, 黄远, 等 . 基于 Self-Attention-BiLSTM 网络的西瓜种苗叶片氮磷钾含量高光谱检测方法[J]. 农业机械学报, 2024, 55(08): 243-252.

[16] 蒋雪松, 计恺豪, 姜洪喆, 等. 深度学习在林果品质无损检测中的研究进展[J]. 农业工程学报, 2024, 40(17): 1-16.

[17] Hou B R, Hu Y H, Zhang P, et al. Potato late blight severity and epidemic period prediction based on Vis/NIR spectroscopy[J]. Agriculture-Basel, 2022, 12(7): 17.

[18] Li Q Y , Hui Y H . Kinetic models of peroxidase activity in potato leaves infected with late blight based on hyperspectral data[J]. International Journal of Agricultural and Biological Engineering, 2019, 12(2): 160-165.

[19] 郭兆华, 文师召, 李思凡, 等. 基于近红外光谱结合数据增强CNN 算法的白芷产地溯源方法[J]. 中国药学杂志, 2024, 59(21): 2022-2029.

[20] Tang Z J, Zhang J Y, Hu M L, et al. Improving streamflow forecasting in semi-arid basins by combining data segmentation and attention-based deep learning[J]. Journal of Hydrology, 2024, 643: 14.

[21] 温廷新, 郭晓赛. 基于特征组合的ECA-TCN光伏发电功率预测模型[J]. 太阳能学报, 2024, 45(12): 94-100.

[22] 赵玉, 陈丽霞, 梁梦姣. 基于LSTM＿TCN模型的降雨型滑坡时间概率预测及气象预警建模[J]. 地质科技通报, 2024, 43(02): 201-214.

[23] 崔笑宁, 苏丹华, 尚维. 基于互联网新闻和时间卷积长短时记忆神经网络的股票指数预测研究[J]. 管理评论, 2024, 36(07): 113-127.

[24] Zhang D D , Chen B A , Zhu H Y , et al. Short-term wind power prediction based on two-layer decomposition and BiTCN-BiLSTM-attention model[J]. Energy, 2023, 285: 19.

[25] Wang W C, Ye F R, Wang Y Y, et al. A singular spectrum analysis-enhanced BiTCN-selfattention model for runoff prediction[J]. Earth Science Informatics, 2025, 18(1): 29.

[26] 刘涛, 刘望, 杨奉源, 等. 基于无人机高光谱影像和机器学习算法的花生生物量估算方法研究[J]. 中国农业大学学报, 2025, 30(03): 206-217.

[27] 汤森林, 张霞, 戚文超, 等. 基于长短时记忆神经网络的葡萄叶面积指数高光谱反演[J]. 遥感信息, 2022, 37(05): 38-44.

基金资助

科技创新2030-“新一代人工智能”重大项目(2021ZD0113603)