Abstract: Farming industry involves classification operation of material. Designs of automatic material classification system is related to pattern recognition and intelligent control techniques. In this paper, the fundamental regular patterns of the classification procedure of agricultural product material were discussed . According to the time restraint of nonintermittent discrete material and the nonseparated characteristic of inherent classification feature of agricultural product material, the concepts of time sufficiency and feature efficiency were proposed . The time sufficiency of a system of material classification must be larger than 0 and the feature efficiency of that should be approached to 1 from positive direction based on principle of minimum dimensions of features. The primary construction and its implementation of multidimensional asynchronous pattern classifier were presented. Online optimization of classification threshold can be finished by a learning program based on Bayes minimum error law or Bayes minimum risk law. According to the idea of manmachine intelligent synthesis, the builtin intelligence system and the extension intelligence system can be established for controlling a material classification system. Human intelligence can be inducted by the extension intelligence system to handle complex cases while the material classification system working. Abovementioned techniques were applied successfully in intelligent classification system for separating tea and stalk. 