Estimating corn leaf chlorophyll content using airborne multispectral imagery and machine learning

Chlorophyll is crucial for photosynthesis and impacts plant growth and yield in crops. Accurate estimation of plant health and fertilizer status is essential for effective nitrogen (N) management in corn. However, crop chlorophyll is primarily quantified using handheld sensors, which is time-consuming, labor-intensive, and of low spatial resolution. This study aimed to evaluate an airborne multispectral imaging system in estimating the chlorophyll content of corn leaves at four vegetative growth stages. Three replicates of 12 nitrogen rates (between 0 and 285 kg ha⁻¹) were applied to corn at the V4 vegetative stage (i.e., with four established leaves). Soil apparent electrical conductivity (EC_a) of all test plots was measured before planting and corn leaf chlorophyll content was measured using a commercial handheld chlorophyll meter at four vegetative stages (V8, V9, V11, and V12). A UAV-based multispectral camera collected imagery at the same time as manual readings. Machine learning models developed based on image features derived from UAV images were used to predict leaf chlorophyll content. Results showed that an epsilon support vector regression model built on imagery data across imagery data collected over four growth stages with a sequential forward feature selection achieved the best performance (R² = 0.87, MAE = 1.80, and RMSE = 2.26 SPAD units). There was no significant difference in the performance of models across the four growth stages. By utilizing the developed model, researchers and growers can effectively map the chlorophyll content of corn leaves at different growth stages, enabling them to make timely and informed management decisions.