Integration of 3D registration methods based on multiple natural features

Current markerless 3D registration methods usually utilize a single specified category of natural features. However, the richness of various natural features is ignored. They cannot meet users´ diversified registration requirements. This article proposes an integration strategy of 3D registration methods based on multiple natural features. Utilizing feature detection and tracking and 3D registration algorithms, taking feature points, rectangle features, natural markers and hand features as four typical natural features, we design a description method of multiple natural features state space. With user interaction, the 3D registration is completed through detecting and tracking corresponding natural features. Experiments demonstrate the applicability and accuracy of the proposed integration strategy of 3D registration methods.