Edge AI-Driven Targeted Fertilizer Dispensing System for Mid-Sized Oil Palm Plantations Using Sensor Fusion

Abstract

Nutrient management in commercial oil palm plantations presents a significant logistical challenge. Manual fertilizer application is labor-intensive and inefficient, while conventional tractor-mounted broadcast spreaders lead to substantial material wastage. These standard machines dispense fertilizer in a 360-degree pattern, causing much of the nutrient load to fall in inter-row spaces where it encourages weed growth rather than crop development. To address these inefficiencies for mid-sized operations, this paper proposes a low-cost, automated fertilizer dispensing system utilizing Edge AI and sensor fusion [4]. The architecture incorporates a Raspberry Pi 4B integrated with an HC-SR04 ultrasonic sensor for proximity measurement and a Pi Camera executing a lightweight YOLOv8 object detection model for real-time oil palm trunk identification [1]. Upon detection of a tree within a 2.5-meter threshold, the system activates a 12V linear actuator to open a targeted chute, delivering a precise dose of granular fertilizer directly to the weeded root basin. The design further includes a mechanical agitator within the hopper to prevent clumping of hygroscopic fertilizers. Initial benchtop evaluations indicate that affordable edge computing devices can achieve the requisite low-latency performance for precision agricultural applications. This system offers a scalable bridge between manual labor and high-cost commercial spreaders, significantly reducing fertilizer waste and operational overhead.