机构地区: 天津大学精密仪器与光电子工程学院精密测试技术及仪器国家重点实验室
出 处: 《计算机辅助设计与图形学学报》 2007年第3期340-345,共6页
摘 要: 利用图像的相关性在不同深度像对应的纹理图中自动选取特征点对;然后根据Hausdorff距离判断对应点对的有效性,并利用这些对应点对获得相应的深度点对,从而计算出2个深度像之间的初始位置变换关系;最后结合一种改进的带纹理的ICP算法实现深度像的精确配准.实验结果证明:该算法不仅适用于自由曲面的深度像匹配,而且适用于规则对称物体的匹配. A method for the coarse registration of range images based on texture features is proposed. By the method, the feature pixel-pairs in two texture images are identified automatically by cross-correlation, and the validity-checking of these pixel-pairs is implemented through Hausdorff distance comparison. The correspondence between the texture image and range image helps to acquire the range point-pairs, and the initial transform of two range images is calculated by the least-square technique. With this initial transform, the fine registration is achieved by Texture ICP algorithm. Not only can this method deal with the registration of two free-form surfaces, but also the objects with highly symmetric shapes. The experiment demonstrates the efficiency of the algorithm.
领 域: [自动化与计算机技术] [自动化与计算机技术]