基于图优化算法的多视图拼接三维可视化技术研究

Multiview Point Registration in 3D Visualization Technology Based on Graph Optimization Algorithm

多视图拼接技术还存在着在大范围区域内连续拼接时误差累积较快、整体精度不足的缺陷,这一问题仅依靠特征点匹配算法难以解决,而SLAM中的图优化方法由于采用了全局优化的策略,具有误差累计增长缓慢的优势,因此可以将特征点提取及匹配算法与图优化方法结合,在时域上进行图像配准,并通过最小二乘法优化从而减少累积误差,提高总体精度。

Registration of multiview point in large space often faces the challenge of rapid error accumulation and insufficient overall accuracy.It is usually difficult to solve this problem only with the help of feature point matching algorithms.The graph optimization method in SLAM has the advantage of slow error accumulation owing to its global optimization strategy.This research therefore combines the feature point extraction and matching algorithm with the graph optimization method to perform image registration in the time domain,and optimizes the results through the least square method to reduce the cumulative error and improves the overall accuracy.