A Kind of 3D Reconstruction Method Based on Affine Transformation

In this paper, a new kind of 3D reconstruction method based on affine transformation is put forward. Firstly, a small amount of matching point pairs with high accuracy are obtained by SIFT algorithm and setting small threshold. Then we obtain affine transformation relationship restriction and polar line restriction matching by these matching point pairs. Contracting the searching region into a small range is useful for improving the speed and accuracy of matching. Moreover, it can reduce the computation. Finally, the 3D points set can be got through space forward intersection by projection coefficient algorithm. In that way, a real 3D model can be reconstructed after Delauny triangulation and texture mapping. Experimental results show that the method can avoid the scarcity of lacking match points and low accuracy which are caused by using only epipolar constraint. Thus, the method can improve the accuracy and efficient of reconstruction significantly.