Moisture content, color and pH value are important quality indexes of beef. Visible and near infrared spectra (NIR) is widely used in meat in recent years due to characteristics such as low cost, rapid and nondestructive. In order to solve the too small of the acquisition area and the poor representation of the spectra obtained by previous detection probe, an efficient ring light source was applied in this study to form a high efficient and stable wide band visible near infrared system. Firstly, the effective spectral data of 400~2450nm were obtained based on the multiple acquisition of fresh beef samples to prepare and analyze the performance of the previous probe and the ring probe, and the result proved that the ring probe had more advantages in stability of which the variation coefficient were less than 10%. Then the spectra of two wavelengths from 61 pieces of beef samples were acquired by the improved system. With no treatment, Savitzky-Golay (S-G), Haar transformation (HT), first derivative (1D), second order derivative (2D), standard normalized variate (SNV), multiplication scatter correction (MSC), detrend (DE) and the combination methods were chosen to pretreat the spectra. The partial least square regression (PLSR) models of color parameters L*, a* and b* of beef were built by the spectra of 400~1050nm, and the models of pH value and moisture content were built with spectra of first wavelength, the second wavelength and the double of wavelength. The result showed that the good PLSR models can be achieve by the spectra of first wavelength without any pretreatment, of which the correlation coefficient of validation set were 0.9603, 0.9616, 0.9367, respectively, and the standard error were 1.3332, 1.1844, 0.6553. And the first wavelength of spectra got better results than the second wavelength, both of which cannot do better than wide band spectra with some pretreatment methods. The correlation coefficients of validation set of water content were 0..9541 and 0.8716, the standard errors were 0.5475 and 0.1272, respectively. The results demonstrated that the prediction model established by the wide band visible near infrared spectrum based on the high efficiency ring probe can realize the accurate and nondestructive testing of the beef quality parameters, and obtain the stable detection results.