3D reconstruction of spatial non cooperative target based on improved traditional algorithm

A new 3D reconstruction strategy is proposed. The strategy obtains the camera internal reference matrix with known camera focal length and field of view angle; Improves the SURF algorithm to add the color information of feature points for feature point extraction and matching; Then takes the imaging time order of the image as a priori information, and solves the 3D sparse point cloud structure of camera position and non-cooperative target by SFM (structure from motion); Finally, CMVS (clustering multi-view stereo) is used to cluster the images., and PMVS (patch-based multi-view stereo) is used to generate the dense 3D point cloud structure by matching, expanding and filtering. The experimental results show that the improved 3D reconstruction strategy is effective to avoid the reconstruction errors caused by spatially non-cooperative targets due to no background texture, structural symmetry, and texture duplication.