Study on Accurate Grading Prediction Model of Tobacco Weather Fleck Based on Internet of Things

Abstract

Abstract Precise prediction of crop disease trends and accurate grading identification of diseases based on information technology represent a significant challenge in the application of IoT devices in agricultural production. Addressing this issue necessitates the development of predictive models for the efficient representation of Internet of Things (IoT) data. In this context, a graded precision forecasting model for Tobacco weather fleck has been developed. Additionally, a graded recognition model has been constructed to identify different levels of disease severity based on the aforementioned grading system. We utilized meteorological data collected from field Internet of Things (IoT) devices and employed the Generalized Additive Model (GAM) to identify factors significantly associated with the occurrence of this disease. The proposed composite model, which leverages the search capability of the Grey Wolf Optimizer (GWO) and the feature extraction advantage of Convolutional Neural Networks (CNN), optimized the Long Short-Term Memory (LSTM) model (GWO-CNN-LSTM) for the best grading prediction effect of tobacco climate spot disease (accuracy rate of 85.46%). The GoogleNet model, optimized with the Convolutional Block Attention Module (CBAM) and based on the Inception-ResNet-v2, achieved the highest accuracy rate for disease grading recognition (92.40%), which was significantly higher than the manual recognition accuracy rate (83.00%; P

Plant phenotyping relevance

タバコ葉の病害状態・重症度を自動予測および画像認識するモデルを開発・比較しており、植物病害表現型の取得・評価手法が研究の中心である。

Code and data availability

The paper describes IoT meteorological data (3198 points), 533 disease records, and 2129 tobacco weather fleck images used for GWO-CNN-LSTM prediction and CBAM-Inception-ResNet-v2 grading models, but no public dataset, image collection, or code repository is deposited. Data availability states the datasets are only 'on

No evidence-backed public reproduction asset is currently recorded.