The essence of crowdsourcing data refers to the existence of a large number of complex and potential junk data in the Internet. How to use effective methods and techniques to collect and use this kind of data of interest has become the center of this writings. This essay adopted the latest technology in the computer field and used the knowledge graph to extract the knowledge representation of crowdsourcing data. In the knowledge graph, knowledge description measures aim to efficiently discover the intrinsic semantic relations between special entities and relationships through a low-dimensional sparse vector representation method. This has important practical significance in application scenarios such as knowledge question answering and information retrieval. However, many existing knowledge description measures ignore remote sensing scene elements and lack geographic knowledge that indicates the change in the scene during use. Aiming at the contradictions in this field, this paper proposed a method for modeling crowdsourcing data based on discrete vectors. This way integrates nuclear power plant scene information into different types of entity vector representations at different levels, which explores the semantic connections between entities and relationships. This paper first described the principle knowledge embodied by knowledge. And then this article introduced the use of traditional artificial intelligence methods to build crowdsourcing data. Last but not least, this paper used the currently recognized semantic web and open knowledge modeling methods to further demonstrate the feasibility of knowledge expression from crowdsourcing data. Finally, the use of global uranium mine knowledge modeling case test showed that this representation measure based on entity discrete vectors can significantly meet the needs of remote sensing scene completion and prediction research for knowledge graphs.