摘要
提出了一种多特征级联目标匹配算法(MFCM)。在先前的研究中提出的基于几何特征的匹配方法复杂度普遍较高,需要改进。MFCM算法首先使用图像特征为每一个特征点建立潜在匹配集,然后进行两级匹配:第一级匹配采用分治思想,利用几何特征建立匹配结果集,并应用投票机制来确保匹配的稳定性;第二级匹配在模板点集和匹配结果集之间建立对应关系。新算法有效提高了几何匹配方法的效率,同时保持了较高的匹配正确率。实验结果表明,MFCM算法可以有效处理大数据量的目标匹配问题。
A new multi-feature cascade object matching （MFCM） method was proposed. Some previous works provided some object matching methods based on geometry feature. However, the complexity of these methods was too high. The key idea the MFCM method is that using the image feature to build the potential matching set for each point and then do the matching process twice： firstly, using the geometry feature to find the exactly corresponding point and add them to the ＂Result Set＂, through the geometry matching process, a vote mechanism was used to ensure the correctness; secondly, building the correspondence between the model point set and result set. The new method has both higher efficiency and correction than other geometry method. The results of experiments show that this new method can deal with the object matching problem with large amount of data.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2012年第1期113-116,共4页
Journal of System Simulation
基金
国家自然科学基金(60970094)
湖南省自然科学基金项目(S2010J504B)
国防科学技术大学预研项目(JC09-06-01)
