摘要
针对用KD-tree实现高维空间点匹配中存在的错误匹配问题进行讨论,分析其存在的原因;接着,使用PCA,根据各维数之间的协方差,求出它们的主成分奉献率,再按主成分奉献率进行维数优先级排序,并在该基础上增加了KD-tree各节点的权重;最后,将改进前后的KD-tree应用于Sift特征点匹配。实验证明,改进后的KD-tree能在保持实时性的前提下,大大提高匹配的准确率。
A specification of the miss matching problem about applying KD-Tree on multi-dimensional point search was given.Principal components analysis was used and the rate of contributions of the main components was calculated.According to the covariance between dimensions,the dimensions were sorted by the rate of contributions and the weight of the nodes on that basis was increased.The improved KD-tree was applied to match the Sift characteristic point.The experiment demonstrates that the matching accuracy is improved greatly in the premise of real time property.
出处
《化工自动化及仪表》
CAS
北大核心
2010年第10期84-87,共4页
Control and Instruments in Chemical Industry
基金
广东省科学中心机器人项目(20071017081958)
