基于Kinect的实物地质标本重建方法研究

Research on the Reconstruction of Physical Geological Specimen Based on Improved Kinect Fusion

为了提升实物地质标本的重建鲁棒性及重建效果,提出一种基于Kinect的三维重建方法。首先,利用Kinect能同时采集目标物体RGB图和深度图的特性,使用联合双边滤波器对深度数据进行优化;然后,对配准后的点云进行全局校准以消除累积误差造成的重建错位;最后,采用局部最优投影法对点云进一步拟合,去除离群噪声,细化点云模型。实验结果表明,该方法能使用较少的输入图像稳定地重建各类实物地质标本,优化重建结果,提高鲁棒性。

In order to enhance the reconstruction robustness and reconstruction effect of physical geological specimens, proposes a 3D reconstruction method based on improved Kinect Fusion algorithm. Firstly, because of the characteristics of RGB and depth images of the target can be acquired simultaneously by Kinect, Joint Bilateral Filter can be used to optimize depth data. Then, carries out the global calibration of point cloud after registration to eliminate the reconstruction error caused by cumulative errors. Finally, conducts the local optimal projection method to further fit the point cloud, removes outlier noise and refines the point cloud model. The experimental results show that the proposed approach can effectively reconstruct all kinds of Geological Specimen with fewer input images, optimize the reconstruction resultsand improve robustness.