摘要
研究虹膜识别问题。针对传统识别算法不能够很好的消除平移,缩放和旋转对于虹膜识别的影响,以及一维log Gabor滤波特征提取方法丢失二维信息的缺陷,导致虹膜识别的正确率低的难题,提出一种改进的虹膜识别算法。首先在虹膜预处理过程中采用粗定位和精定位相结合的虹膜定位方法获取虹膜内外边缘,并用直方图均衡化来增强归一化的虹膜图像,突出虹膜纹理,从而降低光照不均匀性的影响,然后二维log Gabor滤波取代一维log Gabor滤波来提取虹膜纹理特征,克服一维log Gabor滤波丢失二维信息的缺陷,最后对提取的特征进行编码,采用欧式距离进行匹配得到识别结果。通过CASIA虹膜数据库进行了仿真,实验结果表明,相对于当前典型的虹膜识别方法,提高了虹膜识别的正确率,识别速度相应加快,在虹膜识别领域中有着广阔的应用前景。
Researching iris recognition problem.The tradition algorithm cannot eliminate the effects of translation,rotation and scaling on iris recognition,and one-dimensional Gabor loses two-dimensional features,which result in low iris recognition rate.An improved algorithm of iris identification is proposed in this paper.Firstly,the method uses coarse and pure position to obtain iris position,normalizes iris images based on histogram equalization enhance images texture,then replaces one-dimensional iris texture feature extraction with two-dimensional log Gabor filter,which can overcome the defect that one-dimensional log Gabor loses two-dimensional information in extracting and coding the features,and finaly,obtains recognition results from matching.Simulation experiments are carried out with CASIA iris data.The results show that,compared with the typical iris identification methods,this method can improve the accuracy of iris identification,the recognition rate is increased,and the method is prospect in the field of iris recognition.
出处
《计算机仿真》
CSCD
北大核心
2011年第3期313-316,409,共5页
Computer Simulation
关键词
虹膜识别
特征提取
模式匹配
归一化
Iris recognition
Feature extraction
Pattern matching
Normalization
