为了降低基于流形学习的物体姿态估计算法中数据降维算法的复杂度,提高算法的执行速度,提出了一种快速的Isomap 算法. 通过分析Isomap 算法的执行过程发现,计算任意两点间的测地线距离是导致其计算复杂度高的原因之一.基于这一分析,首先假定在空间旋转角度相邻的两幅图像,降维后其对应的数据点在低维流形上也相邻,然后对Isomap 算法中的测地线距离矩阵构造进行优化,优化后不再需要遍历所有数据点,可以大大降低算法的计算复杂度. 实验结果表明:在保证算法效果的前提下与原算法相比,本算法提高了执行速度,且图像序列越长,速度提升越明显,当图像数量达到350 幅时,降维所需时间为原来的13%.%In order to reduce the complexity of the dimensionality reduction in the object pose estimation algorithm based on manifold learning,and to improve the execution speed of the algorithm,a fast Isomap algorithm is proposed.By analyzing the calculation process of Isomap,it is found that calculating the geodesic distance between any two points is one of the reasons for its high computational complexity.Based on this analysis,it is assumed that the two images are adjacent,after dimensionality reduction,the corresponding data points are adjacent on the low-dimensional manifold, and then the Isomap can be improved by optimizing the calculation of the geodesic distance matrix.In this method,it is no longer necessary to traverse all the data points,which can greatly reduce the computational complexity of the algorithm.The experimental results show that by ensuring the performances, the efficiency of the proposed algorithm is improved, and the efficiency of the algorithm is highly improved, especially for long image sequences. When the number of images reached 350, the time cost of the proposed algorithm is 13% of the original method.
展开▼
