An approach of semantic-based focused crawling is proposed in order to use semantic resource efficiently. In this paper, a domain-ontology is used to describe the topic of Web crawling. Lexicon of the keywords list are mapped to ontology, and semantic of words are obtained through mapping. Inference services about assertion set expanding and domain-range relation are defined. The semantic relation among keywords can be inferred by inference services. At the same time, the definition of concept about Web page is given. A semantic computational model is proposed by combining inference services mentioned above. In the end, the order of URLs corresponding to their Web page is decided according to the subsumption of topic concepts. The result show that this approach is advanced in harvest-rate and crawling efficiency and is better than some classical algorithms.