基于特征匹配的无相机姿态三维场景重建方法OA

3D scene reconstruction has significant practical value.However,most reconstruction methods rely on known camera poses,while some pose-free approaches require ordered image inputs.To address the problem of 3D scene reconstruction under pose-free and unordered input conditions,a pose-free reconstruction method based on feature matching is proposed.The method computes image proximity through feature matching,constructs a weighted graph,and determines the sequence of camera pose estimation using a maximum spanning tree algorithm,thereby achieving progressive scene reconstruction.Experiments on the Tanks and Temples dataset demonstrate that the proposed method maintains high reconstruction accuracy and stability under unordered input conditions.The novel view synthesis quality and pose estimation accuracy are comparable to those of CF3DGS under ordered inputs,and significantly superior to its performance with unordered inputs,achieving an average PSNR of 31.409 across multiple test scenes.