A Study On Rule Based Patterns Classification

Pattern recognition is the determination of the identity of a object with the object's characteristics. In ship target recognition on radars, the characteristics of the radar echoes only are not enough to make a reliable result. It's necessary to establish rules from the knowledge of human experts as an assistant, and the rule based pattern classification methods need to be studied.Based on the theory of the expert system, this dissertation focuses on the difficulties of establishing the rule base and the reasoning machine in the real life application of ship target recognition. After analyzing the characteristics of the ship target recognition, it proposed a set of rule-based expressions of the expert knowledge, which refer many mature classification ways to build a knowledge base, and reflect the experience of human experts'. Under the instruction of human experts' experience, a reasoning tree structure has been built in the dissertation, which has used the rules in the knowledge base flexibly to classify the unknown target. The reasoning tree also has the ability of self-training under the direction of the training data.From the theories above, a rule based classifying module has been built. The module works fairly well in the sea surveillance radar experiment. The result demonstrates that this set of rule based classification can work smoothly and effectively. A classifying module can be built rapidly on the basis of these theories, once given enough data.