Using Machine Learning and Hyperspectral Images to Assess Damages to Corn Plant Caused by Glyphosate and to Evaluate Recoverability

Glyphosate is the most widely used herbicide in crop production due to widespread adoption of glyphosate-resistant (GR) crops. However, spray glyphosate onto non-target crops from ground or aerial applications can cause severe injury non-GR corn plants. To evaluate damage plants and recoverability damaged plants, we hyperspectral imaging (HSI) technique field experiments with different application rates. This study investigated spectral characteristic assessed plant glyphosate. Based on HSI image analysis, a variation pattern was observed at 1 week after treatment (WAT), 2 WAT, 3 WAT glyphosate-treated It further found that treated rates equal higher than 0.5X (X = 0.866 kilograms acid equivalents/hectare (kg ae/ha) represents recommended rate for GR corn) would suffer unrecoverable damage. Using Jeffries–Matusita distance as sensitivity criterion, three sensitive bands measured spectra were selected create two indices differentiation band ratio normalization forms, respectively. With indices, recoverable classified an overall accuracy greater 95%. Then, machine learning algorithms (k-nearest neighbors, random forest, support vector machine) respectively combined successive projections algorithm models relate feature The results indicated achieved reasonable accuracy, especially group illustrated potential evaluating herbicides injured using data analysis modeling approaches practical weed management fields.