Image semantic retrieval technology is a research focus in big data era. Data samples and data set of some existing image retrieval technology research is in a relatively independent and closed system, the data volume is relatively small, and the retrieval efficiency also shortcomings. This paper studies how to implement the change from keyword search to contentbased retrieval and then to intelligent retrieval based on semantic, at last to cross the semantic gap of information retrieval. Research results can provide online query capability basing on semantic, narrow the semantic gap data and query requirements, and improve intelligent degree of the information platform.