物料分选过程中的模式识别技术与智能化

Pattern Recognition and Intelligent Control Techniques in Classification Procedure of Material

农产品加工过程中涉及物料分选操作，自动物料分选系统的设计与模式识别与智能化技术相关。该文研究了农产品物料分选过程的基本规律，根据非间歇离散型物料分选过程的时间约束以及物料的固有分类特性呈现不可分离性的特点，提出了时间适应度和特征效率的概念。分选系统的时间适应度应该大于零，特征维数最小化的原则是特征效率从正向逼近于１。介绍了异步多维模式分类器的基本结构和实现方法，根据Ｂａｙｅｓ最小错误分类原则或Ｂａｙｅｓ最小风险分类原则，可通过学习程序实现分类阈值的在线寻优。基于人－机智能综合思想，通过建立内环和外环智能体系来控制分选系统工作，在复杂情况下，外环智能体系可将人类自然智能导入系统帮助工作。上述技术在智能化茶叶分选设备上得到成功应用。

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.