AI-based smart inspection of date fruit to support postharvest artificial ripening decisions

Climate change in major date palm growing regions has delayed fruit ripening on the palm and increased the number of unripe and lower‑quality fruit within harvested clusters. Harvests consequently include fully ripe Tamr fruit, fruit that is marketable after postharvest artificial ripening, and fruit that remains immature or abnormal. This change promotes the need for dependable, high-throughput, non-destructive inspection to support data-driven artificial ripening decisions and contain processing costs. In this study, a machine-learning-based inspection system was developed to evaluate and classify individual dates into three operational groups: Tamr (fully ripe), Biser (artificially ripenable), and low-quality (non-ripenable/abnormal). The system integrates nondestructive multi-sensor measurements, including RGB color, inductive moisture, and projected area (size), with a microcontroller-based control layer and an artificial neural network that fuses sensor signals to infer fruit class and, when enabled, actuate air-jet deflection. Under test conditions, the system achieved a maximum classification efficiency of 96% at a conveyor speed of 1.5 m·s⁻¹, an air pressure of 100 kPa, a 0.10 s opening time for the electronic air valve, and a feeding conveyor speed of 0.30 m·s⁻¹. The highest throughput recorded was 255 kg·h⁻¹ at a conveyor speed of 1.0 m·s⁻¹. These results indicate that AI-enabled, nondestructive inspection offers a high-throughput, cost-effective alternative to manual sorting. Enclosed sensing and automated handling reduce contamination risk, resulting in accurate and actionable classifications that enhance post-harvest productivity and ensure consistent compliance with market specifications.