基于kinect传感器的物体三维重建

随着计算机视觉的发展及深度传感器的出现，基于kinect传感器的物体三维重建技术得到了广泛的研究。针对现有的基于kinect传感器的物体重建算法需要将待重建物体置于转台并随转台转动的局限性，提出了一种通用的基于kinect传感器获取的物体彩色和深度图像完成物体三维重建的方法，对深度图像进行预处理以平滑噪声并获取三维点云，提取彩色图像匹配点并根据彩色图像特征点与三维点云之间的关系得到三维特征点，并通过SVD方法完成点云初始配准，然后利用改进的ICP算法完成点云的精确配准，得到物体表面的三维点云，重建出物体的三维模型。实验结果表明，该方法能有效去除深度图像的噪声，并对各个视角下的点云数据进行精确配准，完成物体的三维重建并得到较好的重建效果。

3 D-object Reconstruction Based on Kinect Sensor

With the development of computer vision and the emergence of the depth sensor, 3D-object modeling based on kineet sensor has been researched extensively. To overcome the limitations of existing 3D-object modeling algorithm based on kineet that objects to be rebuilt should be placed on the turntable and rotated with it, this paper presents a universal 3D-object modeling method based on the color and depth images of kineet. First, the depth image preprocessing is used to smooth the noise, and using the depth image after processing to obtain 3D point cloud. Then, extracting the matching points of color ima- ges, and using the relationship between feature points of color images and 3D points to obtain 3D feature points, the initial transformation matrix is calculated by the method of Singular Value Decomposition （SVD） method. Finlay, the improved ICP algorithm is used to accurately registrant the point cloud and obtains 3D points of the surface of object that 3D-object modeling is accomplished. The experiment results shown that the method proposed in this paper can effectively remove the noise in depth images and accu- rately registrant the point clouds of each perspective to accomplish 3D-object modeling very well.