Abstract: In the field of AI and expert system,knowledge representation is very important.From analyzing conventional knowledge representation this paper presents object-oriented knowledge representation method,which is used in the remote sensing image understanding expert system.Interested class is abstracted in object.Various solving algorithms are distributed within object representation.By passing the message between the objects,the total problem-solving procedure is completed.In the system of object-oriented knowledge representation,there is entity match inference mechanism,and also various solving mechanism envolved.Compared with conventional knowledge representation and inference approach,the object-oriented method is more flexible and effective.The experiment results have shown that the inference based on object-oriented approach is reliable.