融合RGB-D相机和光度立体的三维数字化方法研究

Research on 3D digitizing method fusing RGB-D camera and photometric stereo

近年来随着3D技术的发展,如何实现高效、高精度、低成本的三维数字化成为本领域一个研究热点.本文提出了一种以深度信息为约束,结合光度立体法对深度进行修正的重建方法.首先由RGB-D相机得到的原始深度信息估算出光源和物体表面各点的法矢量,然后构造包含深度信息、法向信息和光滑性约束的代价函数,通过对该代价函数作适当变形,采用最小二乘优化求解出较为理想的被测物体表面三维数字化模型.通过实验可以看出:相较于RGBD相机直接测量,本方法的重建结果在噪声抑制和重建细节上有明显的改善.

With the development of 3D Technology, how to achieve the 3D model eficiently, accurately and cheaply has been a hot topic in this ield. In this paper, a new reconstructionmethod is proposed which is basedondepth information andphotometric stereo method to optimize the reconstruction results. Firstly, the llght source and normal vector informationare estimated using the depth map achieved by RGB-D camera. Then the cost function s composed which includes depth information,normal information and smoothness constraint. Finally,with some modification on the cost function,a satisfied 3D model of measured object can be optimized by least square method. Compared with the traditional method,our methods easier to realize,the overall deformation can be avoided and the details of the measured object can be better preserved. Furthermore, with the help of color photo-metric stereo,our method can realize the 3D reconstruction of dynamic non-rigid objects efficiently. It＇s shown from the experimental that the method proposed can achieve a better result with low noise and more details than the depth data direct from RGB-D camera.