The 3D reconstruction of face model with active structured light and stereo vision fusion

Active structured light and multi-view stereo vision methods reconstruct objects to gain 3D model. However, due to the large scale difference between the two types of point clouds, traditional registration method cannot find corresponding points. This paper presents a 3D reconstruction method fused active structured light and multi-view stereo vision. The precision of the 3D reconstruction is improved by combining point clouds obtained from different sensors. Firstly, a scaled-PCA-ICP(Principal Component Analysis-Iterative Closest Point) algorithm is proposed to do the registration that can overcome the influence of large scale variance. Secondly, because it is difficult for two sets of point clouds to choose their registration center, a recalculated registration center is proposed to realize the point clouds registration again. Finally, the two-view point clouds apply the accurate matching to achieve data fusion. Test on two databases show the effectiveness of this fusion reconstruction method.