基于混合Gabor滤波器与加权中心对称LBP的掌纹识别

Palmprint recognition based on hybrid Gabor filter and weighted central symmetric LBP

掌纹识别是一种比较新颖的生物特征识别技术,提取最佳分类特征一直是掌纹识别研究领域的一个重要方向。掌纹图像纹理特征丰富,但传统方法难以准确将其表征。针对此问题,将固定尺度及自适应多尺度Gabor滤波器结合起来,提出基于混合Gabor滤波器与加权中心对称局部二值模式(weighted center symmetric local binary pattern,WCS-LBP)的掌纹识别方法。首先,利用混合Gabor滤波器对掌纹感兴趣区域进行滤波得到特征图像,并将其串联在掌纹特征空间;然后,使用WCS-LBP提取该空间掌纹特征形成特征向量;最后,通过匹配WCS-LBP直方图序列的相似度来实现分类。在PolyU图库、同济大学图库、IIT-D图库和自建非接触图库中进行实验。结果显示,该算法获得的识别率最高分别可达99.7685%、99.5109%、99.0916%和98.5010%,最低等误率分别为0.7945%、1.2357%、1.6725%和2.3391%,且识别时间都在1 s以内,相比其他传统和流行算法具有优势,显示出方法良好的效果。

Palmprint recognition is a relatively new biometric recognition technology,extracting the optimal classifying features from palmprint always is an important research area in the palmprint recognition field.Palmprint images have rich texture features,but traditional methods are difficult to accurately characterize them.In order to solve this problem,a palmprint recognition method based on hybrid Gabor filter and weighted center symmetric local binary pattern(WCS-LBP)is proposed by combining fixed scale and adaptive multi-scale Gabor filter.Firstly,using the hybrid Gabor filter to extract the region of interest of palmprint to obtain a feature image,and connect it in series in the palmprint feature space.Then,using the WCS-LBP to extract the spatial palmprint features to form a feature vector.Finally,the classification is achieved by matching the similarity of WCS-LBP histogram sequences.Experiments were carried out in PolyU library,Tongji University library,IIT-D library and self-built non-contact library.The results show that the highest recognition rates obtained by this algorithm are 99.7685%,99.5109%,99.0916%and 98.5010%,and the lowest equal error rate rates are 0.7945%,1.2357%,1.6725%and 2.3391%,respectively,and recognition time is within 1 s,which is superior to other traditional and popular algorithms and shows good results.