Milk powder is rich in human body’s five major nutrients, and is one of the main sources of nutrition for infants and young children also. So the nutritional components in milk powder have an important impact on the growth and development of infants. However, excessive sugar content may have adverse effects on the health of infants, except lactose. The chromatography and near infrared spectroscopy detection techniques are difficult to meet the requirements of the rapid and nondestructive testing of milk powder, due to the complex composition of milk powder. Therefore, it is necessary to explore a fast and nondestructive testing method for the content of glucose and sucrose in milk powder. Terahertz wave has fingerprint characteristics for the absorption peaks of different macromolecular substances, which can be used to identify different macromolecular substances. In this paper, the terahertz time domain spectroscopy (THz-TDS) combined with chemometrics method is used to discuss the qualitative and quantitative methods for the determination of glucose and sucrose in milk powder. The experimental device adopts the TAS7500TS terahertz spectroscopy system, and the experimental samples are free-sugar infant milk powder and glucose and sucrose crystals whose purity is greater than 99%, and different concentration of milk powder-glucose, milk powder-sucrose mixture. The terahertz time domain signals of the pressed-slices of pure milk powder, glucose, sucrose and 15 different concentrations of milk powder-glucose, milk powder-sucrose mixture sample, each sample was collected three times and the mean value was taken as the time domain spectral signal. Then the terahertz frequency domain signals of each sample are obtained by Fast Fourier Transform (FFT). The absorption coefficient spectrum and refractive index spectrum of all samples are calculated according to the Dorney optical parameter extraction formula. The partial least squares (PLS) method is used to establish the corresponding quantitative analysis model, based on the absorption coefficient spectra and refractive index spectra of the two mixture samples, respectively, and the ratio of calibration set and prediction set is 2∶1. The experimental results show that there is no obvious absorption peak in the terahertz band, while glucose and sucrose have strong characteristic absorption peaks at the frequencies of 1.45, 1.8, 1.98, 2.7 THz and 1.5, 1.9 and 2.6 THz, respectively. So the glucose and sucrose can be identified according to the THz “fingerprint” characteristic peaks. Meanwhile, the absorption peak of the two mixtures with the concentration varied basically are same as the location of terahertz absorption peak of pure glucose and sucrose, and have stable absorption characteristics. Quantitative analysis of glucose and sucrose in milk powder can be achieved by partial least square (PLS) method based on absorption coefficient spectrum and refractive index spectrum data, and the mixture regression models of glucose and sucrose PLS based on the sample refractive index spectrum are better than the models established based on absorption coefficient spectrum. The former modeling result shows that, the correction set correlation coefficient Rc and root mean square error RMSEC of the PLS regression model of milk powder-glucose mixture are 0.99 and 0.18%, and the prediction set RP and RMSEP are 0.96 and 0.66%. The correlation coefficient Rc and root mean square error RMSEC of the PLS model of sucrose content in the milk powder-sucrose mixture are 0.96 and 0.55%, and the predictive set RP and RMSEP are 0.99 and 0.25%. So the prediction results of glucose and sucrose quantitative analysis models are ideal. The results show that terahertz time-domain spectroscopy can be used in the qualitative and quantitative analysis of glucose and sucrose in milk powder, which provides a reference for the study of adulteration and rapid quality detection of milk powder by using THz-TDS technology.