摘要
针对三维物体重建过程中存在的配准精度差问题,提出一种改进的ICP算法。先利用动态设定的距离函数阈值,再利用高斯曲率相似性,去除错误匹配点对,在迭代结束后给出结果评价准则,提高ICP算法的精度。在使用Kinect相机采集到点云数据后,利用改进的ICP算法进行局部配准,将局部配准的结果应用于全局配准,得到完整的三维物体模型。实验结果表明,该算法配准精度高,能有效地应用于三维物体重建。
For the problem of poor registration accuracy existing in the three-dimensional object reconstruction process ,an im-proved ICP algorithm was proposed .In the algorithm ,the error matching points were removed using dynamically set threshold value of distance function and the similarity of Gaussian curvature ,and the results evaluation criteria was offered after the itera-tion .Therefore the accuracy of the ICP algorithm was improved effectively .After getting point cloud data by using Kinect ca-mera ,the point cloud was registered locally using the improved ICP algorithm .Then the results of local registration were used in the global registration ,and a whole model of the three-dimensional object was generated .Experimental results show the high registration accuracy of the proposed algorithm .The proposed algorithm can be used in the three-dimensional object reconstruc-tion efficiently .
出处
《计算机工程与设计》
CSCD
北大核心
2014年第10期3574-3578,共5页
Computer Engineering and Design
基金
国家自然科学基金项目(61379080)
