In order to vertify the feasibility of the near-infrared to detect the veterinary drug in honey, studying the impact of different background correction methods for linear modeling and nonlinear modeling. Use PLSR to establish the linear model between near-infrared spectroscopy and to be measured, and LSSVR to establish the nonlinear model between near-infrared spectroscopy and to be measured. At the same time, we used Wavelet transform, Multiplicative Scatter Correction, Standard Normalized Variate, First derivative transform, Second derivative transform, Orthogonal Signal Correction and other background correction methods to improve the feasibility of the near-infrared detection.