基于候选匹配区域定位的跨面积手指静脉识别方法

打开文本图片集

中图分类号：TP391文献标志码：A

Cross-area Finger Vein Recognition Method Based on Candidate Matching Region Localization

SHI Xiufeng， CHI Yunhao， YANG Lu

（School of Computer Science andTechnology，Shandong Jianzhu University，Jinan25O101，Shandong，China）

Abstract：To address the cross-area fingervein recognition problemcaused by thearea diference of training and testing images，a cross-area finger vein recognition method based on candidate matching region localization was proposed.The full finger enrolled image was partitioned into multipleblocks，which hada same area withthe smal-area probe image， andthe small-area probe image wassparsely represented using the enrolled image blocks.Theenroled image blocks with non-zero coeficients were seen as the candidate matching regions for the probe image.The finger vein grayscale image and the vein backbone were respectively used to locate the candidate matching regions，and the two located matching region sets were fused forthesecond stage.Local binary patern（LBP）features ofthe probe image and the candidate regionswereextracted，andtheLBPfeaturesof theprobe imageandallthecandidateregions ofone enrolled image were used for Hamming distance based one-to-many matching，and the minimum value of multiple Hamming distances was calculated forrecognitionofthetesting image.Therecognitionperformanceof theproposed methodwasverified through experimentsontwo public finger veindatabases.Theresultsshow thattherecognitionrates ofthe proposed method in Hong Kong Polytechnic University Finger Vein Database，Shandong University Finger Vein Database are94.34% and （20 81.45% respectively，which are better than the four existing finger vein recognition methods，verifying the efectiveness of the method in cross-area finger vein recognition.

Keywords： fingervein recognition；cro-area recognition;candidate matching region localization；sparse representation; vein backbone

生物特征识别技术利用人的行为特征及生理特征识别身份，典型识别技术包括人脸识别、指纹识别、虹膜识别、步态识别和手指静脉识别等，其中手指静脉识别已取得了可观的研究成果[1-3]。（剩余11051字）