The purpose of this experiment was to build a fiber quality prediction model of alfalfa hay in Gansu province by using portable near infrared spectrometer, so as to establish a rapid detection method for alfalfa hay quality. A total of 493 alfalfa hay samples were selected from Gansu province. By measuring the contents of neutral detergent fiber (NDF) and acid detergent fiber (ADF) and calculating the relative feeding value (RFV), all the samples were divided into calibration set samples and prediction set samples by Kennard-Stone algorithm. Then the Monte-Carlo cross validation algorithm was employed to eliminate the outliers in the calibration set. Various pretreatment methods and spectral segments (950 to 1 650 nm) were used to model different quality parameters by partial least squares (PLS) method, and the best modeling method for each index was determined. The results showed that the best pretreatment method for NDF content prediction model was no pretreatment, and the prediction determination coefficient (R2<inf>p</inf>) and relative prediction deviation (RPD) were 0.970 and 3.389, respectively; the best pretreatment method for ADF content prediction model was SNV method, whose R2<inf>p</inf> and RPD were 0.984 and 5.430, respectively; the best pretreatment method for RFV prediction model was SNV method, whose R2<inf>p</inf> and RPD were 0.944 and 2.770, respectively. It can be seen that the prediction accuracy of NDF and ADF contents in alfalfa hay in Gansu province based on portable near infrared spectrometer in this experiment is better, while the prediction accuracy of RFV is slightly worse. © 2025 Chinese Journal of Animal Science and Veterinary Medicine Co., Ltd.. All rights reserved.
Author Keywords
Index Keywords
