|本期目录/Table of Contents|

一种提高人耳特征点识别度的目标区域提取方法(PDF)

《纳米技术与精密工程》[ISSN:1672-6030/CN:12-1351/O3]

期数:
2015年4期
页码:
271-275
栏目:
精密测量
出版日期:
2015-07-15

文章信息/Info

Title:
A Target Area Detection Method for Improving Ear Feature Point Recognition Degree
作者:
蒋景英12 张琪2 张昊2 卢钧胜2 徐可欣2
1.天津市生物医学检测技术与仪器重点实验室,天津300072; 2.天津大学精密仪器与光电子工程学院,天津 300072
Author(s):
Jiang Jingying12 Zhang Qi2 Zhang Hao2 Lu Junsheng2 Xu Kexin2
1.Tianjin Key Laboratory of Biomedical Detecting Techniques and Instruments, Tianjin 300072,China; 2.School of Precision Instrument and Opto-Electronics Engineering, Tianjin University, Tianjin 300072, China
关键词:
人耳特征识别SURF算法图像分割k-means聚类识别度
Keywords:
ear feature recognition speeded up robust features (SURF) algorithm image segmentation k-means clustering recognition degree (RD)
分类号:
TP391.41
DOI:
10.13494/j.npe.20140122
文献标识码:
A
摘要:
对人耳进行特征识别多采用SURF算法,但该算法应用时极易受到图像中非目标区域的干扰,进而影响人耳特征点的检测和匹配准确度.基于目标区域的人耳特征识别算法可以突出目标区,而尽可能地抑制背景区域的影响.针对此问题,提出一种复合图像分割算法—KRM法作为人耳识别的预处理方法,将图像中人耳所在目标区域提取出来.该KRM法分为3步:首先利用k-means聚类算法将图像初步分割为前景目标区域和背景两类;再通过区域生长算法对过度分割的区域进行合并;最后应用形态学腐蚀的方法进行滤波得到人耳所在的目标区域.将KRM目标区域提取和SURF方法联用(简称KRM-SURF算法)应用于50组人耳图像,进行人耳特征点的检测与匹配,实验结果表明,特征点识别度(RD)均值达到0.924,KRM法的使用能极大地提高基于SURF算法的人耳特征识别的准确性.
Abstract:
SURF (speeded up robust features)algorithm is widely used for ear feature matching and recognition. However, the application of the algorithm is usually interfered by non-target areas within the whole image, and the interference would affect the matching and recognition accuracy of ear features. Ear feature recognition algorithm based on the target area can highlight the impact of target area, and suppress the influence of the background area as much as possible. To solve this problem, a combined image segmentation algorithm, i.e. KRM, was introduced as a preprocessing method in this paper to extract the target area from the image. The present KRM algorithm follows three steps: ①The image was preliminarily segmented into foreground target area and background area by using k-means clustering algorithm; ②Region growing method was used to merge the over-segmented areas; ③Morphology erosion filtering method was applied to obtain the final segmented regions. The combination of KRM and SURF (KRM-SURF algorithm) was employed to detect and match feature points in 50 sets of ear images, achieving recognition degree (RD)of up to 0.924. Results show that, based on SURF algorithm, the KRM algorithm can effectively improve the accuracy of ear feature matching and recognition.

参考文献/References

-

备注/Memo

备注/Memo:
收稿日期: 2015-04-07. 基金项目: 国家高技术研究发展计划(863计划)资助项目(2012AA022602). 作者简介: 蒋景英(1972—), 女, 博士, 副教授. 通讯作者: 蒋景英, jingying@tju.edu.cn.
更新日期/Last Update: 2015-09-17