Spatio-Temporal Vision Mamba Network for Early Stress Forecasting and Disease Progression Analysis in Tomato Crops

Environmental stresses and plant diseases critically affect tomato yield and quality, yet traditional static image-based detection methods fail to capture environmental effects and temporal disease progression. The purpose of this study is to create an improved multimodal system of forecasting early stress and analyzing disease progression in tomato plants. The originality of the research is that it unites spatial and temporal learning by means of using a Spatio-Temporal Vision Mamba Network (ST-VMN) that is a combination of spatial feature extraction of Vision Mamba and temporal modeling of environmental data using Bi-directional Long Short-Term Memory (BiLSTM). The suggested framework combines the images of tomato leaves with meteorological time-series information, which allows learning the comprehensive spatio-temporal features to predict the diseases accurately. The experimental analysis on tomato leaf image datasets and environmental data shows that the proposed model has better classification performance in an accuracy of 99.78% with better AUC scores at different stages of disease severity. The model is very effective in detecting the early onset of stress and is very reliable in predicting disease progressions.