The data that the Internet of Things(IOT) produced is fuzzy and enormous, and the Complex Events Processing(CEP) which is the key part of the Internet of Things need to meet the big data's quantity and complexity. In CEP, how to describe the context and infer from the context is the most important, and in this paper, a fuzzy D-S theory and fuzzy ontology based context model is proposed, according to the model, we design a similarity based context reasoning method. The experimental results show that this method can support fuzzy context in CEP and have better performance and accuracy than other methods.