Journal of Agronomy and Field Innovations

Background: Zea mays L. (maize) is the world's most produced cereal crop, with annual global production exceeding 1.15 billion tonnes. Balanced nutrition is critical for optimal productivity, yet nutrient deficiencies remain a major constraint affecting crop yields in both developed and developing nations. Conventional visual diagnosis of nutrient stress is subjective, delayed, and requires expertise, often identifying deficiencies only after substantial yield losses have occurred.
Objective: This study develops and validates a hyperspectral imaging system integrated with machine learning algorithms to enable early, non-destructive detection of individual nutrient deficiencies (nitrogen, phosphorus, potassium, zinc, iron, magnesium) in maize plants.
Methods: A randomized complete block design (RCBD) evaluated eight treatments (control and six individual nutrient-deficient treatments) replicated four times across 2-hectare trial. A pushbroom hyperspectral camera (400–1000 nm, 200 spectral bands) mounted on a UAV platform acquired imagery during mid-vegetative growth stage. Image preprocessing included radiometric correction, noise removal, and spectral normalization. Spectral indices (NDVI, Red Edge NDVI, MCARI, PRI) were extracted and combined with raw reflectance data to train three machine learning models: Random Forest, Support Vector Machine (SVM), and Extreme Gradient Boosting (XGBoost).
Results: XGBoost achieved highest classification accuracy of 94.2% (±2.1%) in distinguishing nutrient deficiencies from control. Individual model performance: Random Forest 89.7%, SVM 87.3%. Sensitive wavelength regions identified: 550–650 nm (nitrogen), 700–750 nm (phosphorus), 800–900 nm (potassium). Spectral indices demonstrated strong correlation (r > 0.85) with leaf chlorophyll content and nutrient concentrations. Early detection capability enabled nutrient stress identification at V6–V8 developmental stages, 21 days before visual symptoms appeared.
Conclusion: Hyperspectral imaging combined with machine learning provides rapid, non-invasive nutrient deficiency detection enabling timely corrective interventions. Integration with UAV platforms offers practical scalability for precision nutrient management and sustainable maize production under climate variability.