Virtual Reality Scene Construction Based on Multimodal Video Scene Segmentation Algorithm

Video scene segmentation has become one of the research hotspots in the video field because of its important role in improving retrieval accuracy, and plays a very important role in the construction of virtual scenes. In order to realize fast and accurate video scene segmentation, this paper proposes a multi-modal video scene segmentation algorithm based on ant colony algorithm. The algorithm extracts the physical features of different modes in key frames based on the idea of multi-modal feature fusion. The similarity between the same modal data and the correlation of different modal data are combined, and the similarity between different lenses is calculated. The lens similarity matrix is constructed and the ant colony algorithm is used to segment the video scene. The experimental data proves that the algorithm has a good segmentation effect on the video scene.