Comparative Study of Spectrometer Sensors for Corn Moisture Content Prediction and Machine Learning Models

Determining the right harvest time is crucial to ensure seed quality, with moisture content as the primary parameter. This study evaluated the ability of portable near-visible infrared (Vis-NIR) spectrometers (AS7265X, C12880MA, and AS7421) combined with partial least squares regression (PLSR) and artificial neural network (ANN) models to accurately predict corn moisture content. A total of 250 samples of corn variety NK 7207 were analyzed using spectral reflectance data processed with multiple scatter correction (MSC), standard normal variate (SNV), and wavelength selection. The C12880MA sensor provided the best results, with the PLSR model achieving a testing accuracy of R² = 0.91 and the ANN model achieving R² = 0.94. The results show that Vis-NIR portable spectrometer can accurately predict moisture content with high reliability as a non-destructive, low-cost, and efficient device. This technology can help farmers determine the optimal harvest time, improve seed quality, and support agricultural productivity.