Development of a sensor-based embedded system for on-the-go variable-rate fertilizer applicator in maize crop

• Developed a real-time, novel sensor-based system for site-specific nitrogen application in maize using NDVI data. • Developed a simplified polynomial algorithm to convert NDVI into real-time nitrogen rates. • Achieved real-time fertilizer modulation with a response time of 3.6–4.5 seconds. • Validated an embedded system achieving R² = 0.978 for NDVI-based fertilizer rate prediction. • Demonstrated a cost-effective VRT solution with 19.18% less N use and improved maize yield. Efficient nutrient management is key to maximizing crop yield and input use efficiency. This study developed a sensor-based embedded system for on-the-go variable-rate nitrogen application in maize using real-time NDVI data from a GreenSeeker sensor. The system enables site-specific fertilization by detecting nitrogen deficiencies and adjusting fertilizer application rates dynamically. The system integrates a GreenSeeker sensor, a microcontroller, and PWM motor driver to control a fluted roller rotational speed based on NDVI data, enabling automated, site-specific nitrogen application. Initial evaluation under laboratory conditions using a custom-built test rig assessed system responsiveness, control accuracy, and motor performance across varying PWM inputs. Field trials during the vegetative stage of maize used real-time NDVI readings to guide fertilizer application at georeferenced locations. Performance was compared with conventional uniform application method. The developed algorithm showed a strong correlation (R² = 0.87, p < 0.001) between NDVI and yield, enabling accurate mid-season nitrogen prediction. The embedded system effectively adjusted metering mechanism's shaft speed in response to NDVI inputs, with speeds ranging from 123 to 328 rpm as NDVI varied from 0.80 to 0.30. Laboratory tests showed a strong linear correlation (R² = 0.978) between fertilizer rate, NDVI, PWM signal and metering speed. The developed system achieved a 19.18% reduction in fertilizer usage without significant yield loss. The response time between NDVI acquisition and actuator adjustment ranged from 3.60 to 4.53 seconds, demonstrating effective real-time performance. Results demonstrate the potential of sensor-based automation to enhance nutrient use efficiency and support sustainable crop management .