An Integrated Vision- and Environment-Aware Decision-Support System for Smart Harvesting

Abstract

Decisions related to harvest timing and crop condition are still largely based on visual inspection and experience in many small- and medium-scale agricultural settings. Such practices make it difficult to respond consistently to changes in fruit maturity or environmental stress. Although digital tools have been proposed to support these tasks, most existing solutions focus on a single aspect, such as ripeness detection, environmental monitoring, or yield estimation— without providing a unified basis for on-field decision making. This paper describes a locally deployable decision-support system that integrates visual assessment of fruit ripeness with environmental context and simple economic reasoning. Fruit images are analyzed to determine ripeness categories and associated confidence levels, which are then used to derive a relative yield potential indicator. In parallel, temperature and humidity measurements are interpreted through rule-based agronomic conditions to identify possible stress or disease-favorable environments. The system favors transparent logic over black-box prediction, allowing users to understand how conclusions are formed. Operation is confined to a local area network using embedded sensing and vision at the edge, ensuring low latency and independence from continuous cloud access. By supporting more informed harvest decisions and reducing avoidable losses, the proposed system contributes toward Sustainable Development Goal 2 (Zero Hunger) and Sustainable Development Goal 12 (Responsible Consumption and Production). The work demonstrates the value of practical integration across vision, sensing, and decision logic in real agricultural contexts.