Qualitative and quantitative methods for analysis of raw materials of Tongren Wuji Baifeng Wan were developed by near infrared spectroscopy(NIRS) in combination with several chemometrics methods. Standard NIR spectral libraries of raw materials were constructed, based on which the unknown materials were verified by quality control(QC) comparison search. Discriminant analysis based on principle component analysis (PCA) and Mahalanobis distance was used in identification of pullus cum Osse Nigko. The spectra of semi-products were used as the standard database for negative semi-product screening by similarity match. Calibration models for assay of main raw materials, excipients, semi-product and finished product were built by partial least squares (PLS) regression , respectively. A revolutionary miniature NIR blend analyzer was used for on-line endpoint prediction in blending process of semi-product based on moving window standard deviation. It is proved that the qualitative methods can distinguish samples accurately while the quantitative methods were precise and reliable. The proposed methods are fast, simple and environmental-friendly, and can be applied to at-spot detection in industrial field by at-line or on-line control.