Abstract: In order to quick, easy and accurately identify the classification of wheat quality, this paper put forward using identification method of near infrared spectral analysis technology combined with BP neural network for classification of wheat. Samples of wheat were carried out a detailed analysis of spectral data in the process of research.First of all, Mahalanob distance was applied on spectral data filtering, which could eliminate abnormal spectrum.And through principal component analysis explained that using near infrared spectroscopy identify the feasibility of the wheat quality classification. In order to improve the performance of the model, SPXY algorithm was used for reasonable division of samples of wheat.Then using first derivative and SNV which were commonly used in data processing method dealt with spectral data ,.which could eliminate irrelevant information and noise impact on wheat spectral data. partial least squares method was used to compress the data, which could reduce the amount of data and save modeling time.Finally,BP neural network as the modeling method is used to establish identification model of wheat quality. Experimental result showed that: the model identification effect is good, the recognition accuracy of strong gluten samples is as high as 94.4%, the identification accuracy of weak gluten samples is as high as 100%, which has realized quickly and accurately classification between strong gluten and weak gluten wheat, which has a very important significance on wheat production, market trading and food processing.