高级检索

基于邻接矩阵更新的误匹配移除方法

Mismatch Removal Method based on Adjacent Matrix Update

  • 摘要: 针对误匹配移除方法不能有效处理约束条件歧义问题, 提出一种基于邻接矩阵更新的误匹配移除方法. 首先利用点对特征(point pair feature, PPF)约束确定匹配之间的兼容性, 构建无向图; 然后计算邻接矩阵和匹配的投票得分, 对匹配进行排序, 取前K个匹配对邻接矩阵进行更新, 并采用退火技术逐渐减少K值; 最终保留高投票得分的匹配. 在UWA3M, 扩增的ICL-NUIM和WHU-TLS数据集上的实验结果表明, 所提方法既有高的召回率又有高的精确度, F分别取得了0.8821, 0.7069和0.8729, 均优于对比方法, 验证了所提方法的有效性

     

    Abstract: Aiming at that the mismatch removal methods cannot effectively handle the ambiguity problem of constraints, this study proposes a novel mismatch removal method based on adjacency matrix updating. Initially, the compatibility between matches is determined using point pair feature (PPF) constraints to construct an undirected graph. Subsequently, the adjacency matrix and voting scores of the matches are computed to rank the matches. The top K matches are selected to update the adjacency matrix, and an annealing technique is employed to gradually reduce the value of K. Ultimately, matches with high voting scores are retained. Experimental results on the UWA3M, augmented ICL-NUIM, and WHU-TLS datasets demonstrate that the proposed method achieves both high recall and precision, with F of 0.882 1, 0.706 9, and 0.872 9, respectively, outperforming comparative methods and validating the effectiveness of the proposed approach.

     

/

返回文章
返回