Recent advances in networking and sensor technologies allow various physical world objects connected to form the Internet of Things (IOT). As more sensor networks are being deployed in agriculture today, there is a vision of integrating different agriculture IT system into the agriculture IOT. The key challenge of such integration is how to deal with semantic heterogeneity of multiple information resources. The paper proposes an ontology-based approach to describe and extract the semantics of agriculture objects and provides a mechanism for sharing and reusing agriculture knowledge to solve the semantic interoperation problem. AgOnt, ontology for the agriculture IOT, is built from agriculture terminologies and the lifecycles including seeds, grains, transportation, storage and consumption. According to this unified meta-model, heterogeneous agriculture data sources can be integrated and accessed seamlessly.