Feed ingredients with high-fiber are widely available and a price competitive source of energy and nutrients for nonruminants. Rapid analysis of detergent fiber content in highfiber feed ingredients, including neutral detergent fiber (NDF) and acid detergent fiber (ADF), contributes to a fine regulation of dietary fiber levels in feed, and the exercise of its nutritional value. Classical Van Soest method for detergent fiber determination is tedious and sensitive to particle size and high cost in the consumption of filter-bag. It is not suitable for feed plant regular detection. The nearinfrared spectroscopy (NIRS) technique is nondestructive, non-pollutive, fast and relatively inexpensive. NIRS methods have been widely adapted in conventional composition analysis of feed. Feed ingredients with highfiber normally derives from a wealth of sources, like byproducts of grain plant. Single species NIRS model is hardly satisfied the demand of various ingredients in one feed plant. Researchers devoted to developing a unified NIRS model to determination of NDF and ANF for different feed ingredients with highfiber. A NIRS model for NDF and a NIRS model for ADF were built for analyzing six kinds of feed ingredients, including maize DDGS, soybean hull, wheat bran, alfalfa pellet, shotcrete corn bran, and beet pulp particles. Totally 327 samples were collected from China, including 75 maize DDGS, 48 soybean hull, 48 wheat bran, 48 alfalfa pellet, 59 shotcrete corn bran, and 49 beet pulp particles. NDF and ADF were obtained by using Van Soest methods. Fouriertransform nearinfrared spectrometer collected nearinfrared diffuse reflectance spectroscopy with spectral range of 4000~10000cm-1, at the resolution of 16cm-1 and 64 times scanning per scan. Calibration models were constructed by using partial least squares method, combined with derivative processing, multiplicative scatter correction and so on. Results showed that NDF and ADF distribution of six kinds of feed ingredients were respectively 21.20% ~ 65.28%, 6.40%~48.31%. For NDF model: determination coefficients of calibration and verification were 0.971 and 0.963 respectively; RMSEs of calibration and verification were 1.68 and 1.82 respectively; the relative percent deviation was 5.2, and optimal pretreatment method was the second derivative. For ADF model: determination coefficients of calibration and verification were 0.990 and 0.985 respectively; RMSEs of calibration and verification were 1.32 and 1.63 respectively; the relative percent deviation was 8.2 and the optimal pretreatment method was the first derivative with multiplicative scatter correction. These models are appropriate to a variety of high-fiber feed ingredients for the determination of NDF and ADF in feed plant.