The video of civil engineering supervision provides an effective method to improve the quality of civil engineering supervision. This paper first took the video retrieval of civil engineering supervision as the research object, established the semantics of civil engineering supervision video, and regulated the semantic division of civil engineering supervision video. It trained some related Chinese entries in Wikipedia and several civil engineering supervision entries in the domain on word vectors, and used these word vector data to train the labeled data items in order to provide the word vector data with semantic for the R-tree in the video of civil engineering supervision. Then it studied the nodes splitting based on spectral clustering and the proposed the corresponding algorithm, and also proposed the R-tree node retrieval algorithm based on word vector. Finally, the practical experimental results show that the semantic proposed is able to represent the video in the civil engineering supervision, and the proposed algorithms can effectively improve the indexing speed and recall rate of civil engineering supervision video.

[2019-06-05] Printed Article /doc/2018.03.0133/v2