Data-Driven Precision Harvest Planning via Visual Growth Forecasting

Abstract

While Artificial Intelligence (AI) techniques have been widely applied to problem-solving, their true transformative power lies in enhancing people’s practice. For instance, in greenhouse agriculture, data analytics can enhance crop monitoring and resource planning; computer vision can support detecting crops and their conditions; and machine learning methods can foster tasks such as quality assessment and yield prediction. However, these technologies are often deployed only in isolated stages of cultivation, highlighting the absence of an integrated pipeline that connects them to what is desired in real-world practices. To bridge this gap, we propose CROPi–Computer vision for Resource Optimization and Precision harvest using Image data. CROPi is developed with four core functionalities: i) we propose and justify using hue as an objective redefinition of fruit ripeness; ii) we parameterize the ripeness development over time by analyzing growth monitoring images; iii) we develop predictive models to forecast the growth of new fruits; and iv) we optimize fruit-level harvest plans by aligning predicted growths with sequential demands. We demonstrate this approach using strawberries as the use case. By integrating insights from data mining and machine learning into harvest planning, CROPi lays the foundation for next-generation precision agriculture, ensuring efficiency, sustainability, and improved yield management.