中国科学院自动化研究所机构知识库

The application of trauma and hemorrhage medical treatment training has generated higher and higher requirements for the authenticity of the medical model. With the increasing integration of mixed reality technology and the medical field, multi-view-based 3D reconstruction technology can generate trauma medical object models not limited to specific targets, with low cost and high precision. However, the current reconstruction methods have weak points such as the inability to build the object surface beyond the direct view of humans, the uneven distribution of reconstructed point clouds, many holes, and reconstructed triangles. And it is timeconsuming and laborious to fill and modify manually and requires highly skilled modeling expertise. Therefore, the models generated by traditional 3D reconstruction methods cannot meet the needs of interaction and touch in mixed reality applications. In order to solve these problems, we propose a new model generation method based on multi-view and point cloud processing, which enables the model to reduce the computational cost under the premise of meeting the need for direct interaction and touch in mixed reality applications. Finally, based on the reconstruction method proposed in this paper, a new medical trauma hemorrhage control and rescue training system is designed and implemented, which proves the effectiveness of the method proposed in this paper.