Under the background of globalization, the speed of information dissemination is accelerated, the spatial distance is gradually narrowed, the types and quantities of information are becoming more and more abundant, and the data presents diverse characteristics, so big data thinking emerges as the times require. “Dance” is a very classic art form. With the development of science and technology, people begin to think about how to show dance in a richer form. In this paper, through optical motion capture, firstly, optical motion capture is used as the representation of motion content, which effectively reflects the locality, rhythm, and subdivision of dance movement, and provides the basis for the realization of national dance motion data retrieval. Then, the dance posture analysis method based on 21 key nodes and 7 feature-matching planes of the feature plane vector is used. Finally, the feedback correlation mechanism is used to capture the query interest of users, Improve the retrieval performance, and under the verification of a large number of experimental data, it is verified that this method has good retrieval performance and can realize dance movement data retrieval.