Research on forage hyperspectral image recognition based on F-SVD and XGBoost

Aiming at the high time complexity and poor accuracy of traditional SVD in hyperspectral recognition. we proposed F-SVD, which introduces latent factors(F) into decomposition strategy uses correlation between variable original to improve singular matrix. Firstly, used F-SVD reduce dimension visible-near infrared image, consequently designed a forage recognition model based on XGBoost. When test set sets 40%, OA F-SVD-XGBoost is 91.67%, takes 0.601s. Compared with FA-XGBoost SVD-XGBoost, increases 1.98% 1.67%, consumption decreases 1.369s 0.522s, respectively. The results show that our not only effectively extracts essential features improves classification, but also has faster processing speed, so can efficiently quickly realize identification images.