Research on Tomato Quality Prediction and Fish Disease Diagnosis Based on Multimodal Data

Abstract

We propose a multimodal data-driven scheme for intelligent tomato quality prediction and fish disease diagnosis to advance smart protected agriculture and modern aquaculture. Integrating multi-source environmental, visual and managerial data, we construct a full-cycle tomato dataset and develop AI prediction and decision models to address the insufficient digitization and experience-dependent management in traditional tomato cultivation. Targeting aquaculture complexity, this work builds a intelligent monitoring system combining underwater robots, multimodal large models and professional knowledge bases. A lightweight detection model integrated with Retrieval-Augmented Generation mitigates model hallucinations and improves diagnostic credibility, while optimized underwater communication protocols guarantee stable data transmission in complex water environments. This study also realizes cross-scenario technology migration from terrestrial planting to aquaculture, achieving accurate, low-cost underwater perception and advanced fish disease early warning. The proposed solutions provide effective technical support for agricultural and aquacultural intelligent upgrading.