摘要
非刚性点集配准的关键是找到点集之间的正确对应关系。传统点集配准方法通常将特征点的全局空间距离作为判别准则,而未考虑点集的邻域结构信息,容易产生误匹配,为此,文中提出了一种基于邻域结构和驱动力的非刚性点集配准算法。首先,在一致性点漂移(CPD)算法的基础上,提出了一种局部距离计算方法,并将其与空间距离相结合,有助于提高匹配精度;此外,对传统形状上下文方法进行了改进,构建了一种新的驱动力准则,以在初始配准过程中提高搜索速度,在后期减小配准误差;最后,采用期望最大化(EM)算法迭代求解各点对的对应关系。在常用国际点集数据集上的仿真实验结果表明,在非刚性变形、噪声、异常点和遮挡等情况下,文中算法比经典算法具有更高的鲁棒性,匹配准确率也更高,并且对真实图像可以获得比较理想的配准效果。
The key to non-rigid point sets registration is to find the correct correspondence between point set.Traditional point set registration methods usually take the global spatial distance of feature points as the criterion but ignore the neighborhood structure information,leading to mismatching.To solve this problem,this paper proposed a non-rigid point sets registration algorithm which is based on neighborhood structure and driving force criterion.Firstly,a method was proposed to calculate the local mixing distance based on the Coherent Point Drift(CPD)algorithm and it was combined with the original space distance to improve the matching precision.Besides,a new driving force criterion is constructed based on the improved shape context,aiming to improve the searching speed in the original matching process and decrease the matching error in the later process.Finally,the correspondence of each point pair was obtained through the Expectation Maximization(EM)algorithm.Simulations were carried out with commonly-used international point set datasets and the results demonstrate that the proposed method has better robustness and accuracy that traditional methods under the conditions of non-rigid deformations,noises,outliers and occlusions.Moreover,the proposed algorithm can achieve ideal registration results on the real images.
作者
何凯
刘志国
李大双
赵岩
HE Kai;LIU Zhiguo;LI Dashuang;ZHAO Yan(School of Electrical and Information Engineering, Tianjin University, Tianjin 300072, China)
出处
《华南理工大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2022年第4期73-80,共8页
Journal of South China University of Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(62171314)。
关键词
点集配准
非刚性配准
邻域结构
驱动力准则
高斯混合模型
point set registration
non-rigid registration
neighborhood structure
driving force criterion
Gaussian mixture model
