摘要
为了提高图像识别的准确性,提出一种基于可拓学的E-SURF图像识别算法。首先,利用可拓学的发散分析原理为SURF算法增加拓展方向,然后,利用可拓变换中的置换变换改进特征点匹配算法,使得当利用特征点主方向产生的描述子不能找到匹配对时,用置换变换使拓展方向为特征点主方向,然后计算新的描述子,再次进行匹配,从而解决特征点对匹配过程中的矛盾问题。实验证明,该算法能够提高常见变换下的图像识别的准确性。
In order to improve the accuracy of image recognition, the paper proposes an E-SURF image recognition algorithm based on extenies. Firstly, following the divergence analysis principle of extenics, it adds extension orientation for SURF algorithm; secondly, by utilizing displacement transformation of extenics transformation it improves the feature points matching algorithm so that when the descriptor generated by feature points main orientation fails to find its matchers, displacement transformation is used to take the extension orientation as feature points main orientation; thirdly, the new descriptor is calculated and re-matched so that the conflict during the matching process of feature points is solved. Experiment proves that the algorithm can improve the accuracy of image recognition with common transformations.
出处
《计算机应用与软件》
CSCD
北大核心
2014年第7期221-225,共5页
Computer Applications and Software
基金
河南省科技厅科技攻关项目(122300410323) 。
关键词
图像识别
可拓学
发散分析
置换变换
Image recognition
Extenics Divergence analysis
Displacement transformation
