A New Approach to 3D Modeling Using Structured Light Pattern
A New Approach to 3D Modeling Using Structured Light Pattern
复制标题
DOI：
10.1109/ictta.2008.4530071
复制
发表时间：
2008-04
期刊：
影响因子：
--
通讯作者：
M. A. Tehrani;A. Saghaeian;O.R. Mohajerani
中科院分区：
文献类型：
--
作者：
M. A. Tehrani;A. Saghaeian;O.R. Mohajerani
随着三维曲面造型的广泛应用，人们对三维曲面造型的需求越来越大，新的技术也在不断发展和优化以满足这种需求。图像测距技术主要分为主动测距和被动测距。在主动投影仪测距技术中，采用单摄像机和高聚焦性能的图案投影仪对所拍摄的图像进行可靠的测量。被动投影仪测距技术基于寻找从同一物体表面拍摄的立体图像之间的对应关系。这两种技术都使用投影仪来投射结构光图案。使用结构光的优点之一是向对象添加一些额外的特征，这些特征可以包括以这种方式重建的本质上没有特征的对象。在本研究中，结合这两种技术的显著优势，实现了一种实用的三维建模技术。首先，设计了一种彩色狭缝图案来唯一地标记和精确地提取沿狭缝边缘的图像点。任何图像点的标签是根据相邻狭缝上具有最大色调差异的颜色序列的下一个和前一个狭缝的色调值来分配的。然后，沿着立体图像的核线提取狭缝的边缘点，并同时进行标记，将这两种策略结合起来，增强了狭缝出现的唯一性和精确性。第一种策略是消除背景，从而减少表面颜色的影响。第二种策略是对光缝进行颜色排序，简化了相应核极点之间正确点的匹配。最后，将所设计的缝隙图案应用于平面墙体、工业部件和人脸。计算了三个不同表面的颜色点云，计算出的平面上的点代表了设计图案的可接受精度。对工业零件的点提取和标注算法的稳定性进行了测试，人脸的点云显示了该算法能够获得场景的真实颜色并将其分配给点云的能力。因此，提出的缝隙方向图降低了成像和点提取噪声的影响。
As the need for using and developing three dimensional (3D) surface modeling grows increasingly due to its wide range of applications, new techniques are developed and optimized to fulfill the requirements. The two main image ranging techniques refer to the active and passive approaches. In the active projector ranging technique, a single camera and a pattern projector with high focusing performance are used to obtain reliable measurement on the images taken. The passive projector ranging technique is based on finding correspondence between stereo images taken from the same object surface. Both techniques use a light projector which projects a structured light pattern. One of the advantages of using structured light is adding some extra features to the object which can include inherently featureless objects to be reconstructed in this way. In this study, the significant advantages of the two techniques are combined to implement a practical 3D modeling technique. First, a color slit pattern has been designed to uniquely labeling and precisely extracting the image points along slits edges. The label of any image points is assigned according to hue value of next and previous slits where the sequence of colors has maximum hue value differences on adjacent slits. Then, edge points of slits are extracted along the epipolar lines of stereo images and labeled simultaneously.Applying two strategies together, we've enhanced the unique and precise occurrences of the slits. The first strategy was background elimination which reduced the effects of surface color as a result. The second strategy was a color ordering for light slits which simplified matching of a correct point among corresponding epipolar points. Finally, the designed slit pattern has been applied using a flat wall, an industrial part and a human face. The color point clouds of the three different surfaces have been calculated and the calculated points on the flat wall have represented an acceptable accuracy of the designed pattern. The stability of the point extraction and labeling algorithm has been tested for an industrial part and the point cloud of the human face shows the capability of the proposed algorithm to get and assign the true color of the scene to the point cloud. Therefore, the proposed slit pattern has reduced the effects of imaging and point extraction noises.