The application of near-infrared spectroscopy technology for the rapid and accurate identification of attributes such as the origin and quality grade of agricultural products can play an important role in the acquisition and processing of agricultural products. However, the practical application of near-infrared technology for qualitative identification is rare because of certain problems related to the reliable of the model. Taking tobacco leaf samples with different parts (grades) in Sichuan Province, China as an example, this paper discusses a method for determining whether it is reliable to a grade identification model according to the main chemical composition, near-infrared spectroscopy and qualitative discrimination results; and a method for analyzing tobacco leaves grade characteristics according to main chemical composition and near-infrared spectroscopy. Within a certain ecological production area, a reliable leaf grade recognition model can be established, and the consistency of the material information base and the model identification results can verify the reliability of the model. By exploring the chemical composition and spectral characteristics, the possible grade characteristics of tobacco leaves in Sichuan Province are analyzed: the upper tobacco leaves have low total sugar, high nicotine, high total nitrogen, high cellulose and high amide content; the middle have high total sugar, middle nicotine, middle total nitrogen, middle cellulose and middle amide content; the lower has a high total sugar, low nicotine, low total nitrogen, low cellulose and low amide content. The methods used herein, such as the method for determining the reliability of qualitative models based on the material information basis and the method for analyzing the characteristics of tobacco grades based on the chemical composition and spectral characteristics, have reference value for modeling the attributes and chemical composition analysis of other agricultural products.