The Research Of Semantic Nets Based On Mental Imagery

Knowledge representation is one of the major study subjects in the academic fields of man-made intelligence, there into the semantic nets structurizedly represent the mutual relationship of things, which has been widely used in the knowledge representation system. Semantic Network, as the tool of knowledge representation in artificial intelligence domain, represents knowledge by directed graph, adapt to the logical reasoning. The SNetL(Semantic Network Language),which designed and developed by professor Xia Youming, is the formalized language of Semantic Network, The core is describing knowledge in the form of triple, realizing the separating idea by tab expression. The relations are handled as nodes to establish the relationships between the relations. And to describe the meanings expressed by natural language. This can provide better question putting and answering ability.Mental imagery methodology simulates the manner of human habitual thinking and keeps questions pace and answer space consistent as far as possible, thus it has distinctive superiority in knowledge representation.This paper tries to combining the advantages of semantic nets and mental imagery methodology, and presents a new method of relationship model representation for computer vision: semantic nets of mental imagery methodology. This paper, in terms of theory, sets for the each expects of mental imagery methodology and analyses its application in computer vision. The method of object-oriented describes the thing being studyingdoubly, viz. static describing and dynamic describing. The first expresses the object's classified attribute, the latter expresses the object's action attribute. The two are interactional and also interdependent. The static describing expresses the object's state;The dynamic describing divides into two parts. one is the operation to modify attribute state of self the other one is the operation to print the result. The "object" is a entity composed of the data and the code to operate the data, it has special attribute and action mode.The mental imagery descriptions of relating node is included in the node of semantic nets of mental imagery methodology, and contains attributes and operation. It can greatly reduce the quantity and variety of node and semantic relationship, simplify the semantic nets and lead to simple structure, clear level and convenient reasoning.The default value is particularly introduced in the mental imagery description of the class node in semantic nets of mental imagery methodology. The particularity of special examples can be maintained by this default mechanism in reasoning processing. That method of non-monotonic reasoning matches the manner of human's thinking.The operations are introduced in to the mental imagery description to the node of class and object, so that the semantic nets of mental imagery methodology can reason new knowledge according to the current knowledge through operation and strengthen it self's learning ability.In the description of each node of semantic nets of mental imagery methodology, it is specially indicated that the attributes can indicate the movement manners, which is neglected in the normal knowledgerepresentation. The movement attributes can help the semantic nets mental imagery methodology give expression to the moving scene more if effectively.This paper not only interprets the definition and description of semantic nets of mental imagery methodology, but also analyses inheritance, matching and the other concerned issues. Moreover it adopts the C++ object-oriented methodology to construct a simple scene and analyzes the realizational methods and technology of semantic nets of mental imagery methodology.