Construction of an industrial detection model for tomato organic acids based on low-rank adaptation and spatial feature fusion

Tomato (Solanum lycopersicum L.) is not only a globally important crop for human consumption but also a potential source of industrially valuable organic acids. Organic acids such as citric and malic acid have widespread applications in bio-based chemical synthesis, fermentation, biodegradable plastics, and green solvents. In this study, we develop a novel non-destructive framework combining Low-Rank Adaptation (LoRA) with Adaptive Spatial Feature Fusion Network (ASFFN) to accurately predict organic acid content using hyperspectral imaging (HSI). The model integrates advanced feature fusion strategies and robust outlier detection to improve generalizability across tomato varieties. Comparative experiments demonstrate the superior performance of the proposed LoRA-ASFFN over conventional CNN and PLSR baselines. The model enables rapid and precise identification of tomatoes with high organic acid concentrations, providing an efficient pathway for industrial processing and extraction. This research contributes to advancing bio-based chemical supply chains and improving the economic value of tomato crops through data-driven, precision screening techniques.