A digital framework for smart greenhouse tomato cultivation

To address the issues of lagging management, low informatization, and high reliance on manual labor in traditional greenhouse tomato cultivation, this study constructed an integrated digital system by converging the reduced costs (IoT), big data analytics, artificial intelligence (AI), and digital twin technologies. The core of the system involves deploying a sensor network and machine vision for real-time data acquisition, identifying key environmental factors via Principal Component Analysis for advanced prediction with optimized LSTM models, developing a lightweight convolutional neural network with a texture attention mechanism for intelligent disease identification, and establishing an intelligent cloud-based decision and control system integrated with a digital twin. An application demonstration in a Guangdong greenhouse showed that the system effectively optimized the growth environment, shortening the tomato growth cycle by 10%-15%, increasing yield by 20%-30%, and achieving approximately 60% water savings. The developed models demonstrated excellent performance: the CO 2 prediction model achieved an RMSE of 0.0194, and the disease identification model attained a mean Average Precision (mAP) of 96.5%. The proposed system provides a comprehensive technical solution and offers significant value for promoting the practical development and industrial transformation of precision, smart agriculture.