高识别率的鲁棒ART快速图像识别算法  

On ART Based Strong Noise Robust Fast Image Matching Algorithm

作　　者：刘内美[1] 

机构地区：[1]内江职业技术学院会计系,四川内江641000

出　　处：《西南师范大学学报（自然科学版）》2016年第8期106-113,共8页Journal of Southwest China Normal University(Natural Science Edition)

基　　金：四川省科技厅支撑计划项目(2013GZ0030)

摘　　要：角放射变换(ART)形状描述符的图像匹配算法识别性能不佳且易受噪声影响,对此,提出了一种高识别率、具有强噪声鲁棒性的改进的ART图像匹配算法.首先,利用ART径向基和角度基的递推关系,快速计算出ART的转换系数;然后,对ART矩阵函数求导,并计算所有局部极小值;最终求得全局最小值.实验结果表明,本算法具有图像识别率高、强噪声鲁棒性以及低时间复杂度等优点,同时开发了快速匹配算法.The image matching performance of Angular radial transform （ART） shape descriptor is not good and be easily influenced by noise, to overcome these shortcomings, proposed a high recognize per- formance, strong noise robust improved ART image matching schema. Firstly, the recurrence relations of ART radial and angular kernel functions has been used, and the ART coefficient fast computed. Secondly, the roots have been computed for the derived function of the ART moments. Lastly, the global of mini- mum value has been computed, and the image matching schema realized. Experimental results show that the proposed schema have the features of high image recognize performance, strong noise robust and low computational complexity.

关 键 词：角放射变换 区域形状描述符 全局最小值 相似性检测 几何不变性 数值稳定 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]