高光谱成像
人工智能
模式识别(心理学)
数学
卷积神经网络
集合(抽象数据类型)
像素
数据集
遥感
支持向量机
计算机科学
偏最小二乘回归
植物病害
机器视觉
人工神经网络
可视化
作者
Yongxian Wang,Ruofei Liu,Yuqi Wang,Zeju Li,Yongqiang Zheng,Chengming Yu,Chengli Wang,Hongjian Zhang,Jinxing Wang
标识
DOI:10.1016/j.atech.2026.102706
摘要
Apple leaf health is vital for photosynthesis and fruit yield. Ring rot, an important foliar disease, damages tissues, reduces chlorophyll, and impairs photosynthesis, thereby affecting tree growth and fruit quality. To achieve early diagnosis of ring rot and prediction of SPAD values in apple leaves, this study proposes a method based on hyperspectral imaging combined with a Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) hybrid model. A hyperspectral imaging system (400–1000 nm) was used to collect apple leaf hyperspectral data across five infection stages over two consecutive years. Characteristic bands associated with disease stages and SPAD values were selected using competitive adaptive reweighted sampling (CARS) and the successive projections algorithm (SPA), respectively, and multiple machine learning and deep learning models were constructed for comparison. The results showed that the proposed CNN-LSTM model achieved the best performance for both tasks. For early disease diagnosis, the prediction set yielded a P P of 0.9429, a K P of 0.925, and an F1 P of 0.941. For SPAD prediction, the prediction set achieved an R P 2 of 0.8868, an RMSEP of 2.4103, and an RPD of 2.9872. Furthermore, the SPAD prediction model was applied to pixel-wise analysis of hyperspectral images, enabling spatial visualization of the distribution of predicted SPAD values in apple leaves. Overall, the findings demonstrate the feasibility of hyperspectral imaging combined with a CNN-LSTM hybrid model for early diagnosis of ring rot and physiological status assessment in apple leaves, providing technical support for disease monitoring and health evaluation in apple orchards.
科研通智能强力驱动
Strongly Powered by AbleSci AI