Underwater 3D object reconstruction with multiple views in video stream via structure from motion

Underwater 3D object reconstruction is one of the most essential and fundamental tasks in ocean investigations. In this paper, we try to capture the inherent geometrical variation of 3D objects at multiple visual angels with the underwater vehicles and develop a novel underwater 3D object reconstruction model for the continuous video stream. Image enhancement will be first taken into consideration by guided image filtering to acquire clearer images for object tracking which can be completed by the particle filter. And then, we use scale invariant feature transform (SIFT) and random sample consensus (RANSAC) to detect and correspond the features of objects we have tracked. From these correspondences, the 3D point cloud reconstruction will be finally developed with the help of structure from motion (SFM) and patches-based multi-view stereo (PMVS). It has been shown in the simulation experiments that the developed scheme of this paper achieves consistent performance improvements over 3D object reconstruction for underwater video stream in multiple views.