Currently expert system for plant protection has the problem of poor portability and expensiveness, so a cotton diseases and insect pest diagnosis system based on image rules was developed on the Android smartphone. Binary retrieve rules were used to construct decision tree, object oriented programming was used to encapsulate binary logic classification model, rules and appropriate photographs, consequently, knowledge represented by graphics achieved. The system has two diagnosis methods of image retrieval diagnosis and binary tree retrieval diagnosis with image rules. It featured human-computer interaction on actual image of typical characteristic facts in the field. The reasoning process was realized by visualization. The system was benefit for the agricultural production in practice with excellent features as portable, practicability, friendly interface, information of picture and data, and unlimited with networks. The diagnosis accuracy was above 95%.