Research on Maize Variety Identification Based on Hyperspectral Imaging Technology

Maize, as a globally vital food and economic crop, requires rapid and accurate variety identification to ensure food security. This study developed an innovative nondestructive discrimination method by integrating hyperspectral imaging (400–1000 nm) with machine learning algorithms. Hyperspectral images of 1200 maize seeds from six varieties were acquired and preprocessed using mean centering (MC) to eliminate baseline drift. Principal component analysis (PCA) and successive projections algorithm (SPA) were employed for feature wavelength selection. Three machine learning algorithms—backpropagation neural network (BP‐NN), random forest (RF), and support vector machine with radial basis function (SVM‐RBF)—were systematically compared. The results demonstrated that the RF model with PCA feature selection (MC‐PCA‐RF) achieved optimal performance, with a test set accuracy of 92.33% ( p < 0.05), significantly outperforming traditional approaches such as manual inspection, chemical analysis, and molecular markers. This study proposes a practical integration of existing preprocessing and feature extraction methods, forming an effective pipeline for maize variety identification, which significantly improves classification accuracy and operational feasibility in precision agriculture.