3D Model retrieval technology based on 3D reconstruction

In order to overcome the problem that the low efficiency of current 2D target object recognition system and the process of target object recognition based 3D reconstruction is complex. The paper proposes a 3D model retrieval technology based on 3D reconstruction. The system use a new automatically select 2D feature points technology based Horn-Schunck flow algorithm and canny operator; then use 3D cloud points reconstruction technology based on Expectation-Maximization algorithm (EM) and Structure From Motion algorithm (SFM) reconstruct 3D cloud points of target object; then use a new technology rapid extract Extended Gaussian Image (EGI) based on Crust algorithm to complete matching and retrieving target object, at last prove the system is efficiency and accuracy.