CAJ | 학술논문

由于嵌入式身份识别装置数据处理能力有限，提出一种旨在降低数据处理量，提高数据处理速度，能够在低分辨率图像下保持较高识别性能的手三模态融合识别方法。首先根据手形识别速度快，识别率较低的特点，将登录对象划分为三类，即第一类身份确认，第二类身份否决，第三类身份待定。然后，根据掌脉和掌纹识别准确性高、识别速度慢的特点，仅将处于第三类的登录对象进行低分辨率掌纹掌脉图像的数据融合识别，并将此作为第三类对象的最终识别结果。在自主研制的非接触低分辨率三模态图像采集装置上的实验结果表明，融合后的识别率与单模态相比可提高2.9%到7.1%，与其它融合方法相比提高0.66%到1.89%；手形识别后需要进行掌纹掌脉融合识别的数据量占总数据量的21.56%，数据处理时间约为完全进行掌纹掌脉融合识别所需时间的27.89%，约为其它方法处理时间的15.48%到20.43%，速度提升比较明显，应用于嵌入式数据处理系统具有显著优势。
유우감입식신빈식별장치수거처리능력유한，제출일충지재강저수거처리량，제고수거처리속도，능구재저분변솔도상하보지교고식별성능적수삼모태융합식별방법。수선근거수형식별속도쾌，식별솔교저적특점，장등록대상화분위삼류，즉제일류신빈학인，제이류신빈부결，제삼류신빈대정。연후，근거장맥화장문식별준학성고、식별속도만적특점，부장처우제삼류적등록대상진행저분변솔장문장맥도상적수거융합식별，병장차작위제삼류대상적최종식별결과。재자주연제적비접촉저분변솔삼모태도상채집장치상적실험결과표명，융합후적식별솔여단모태상비가제고2.9%도7.1%，여기타융합방법상비제고0.66%도1.89%；수형식별후수요진행장문장맥융합식별적수거량점총수거량적21.56%，수거처리시간약위완전진행장문장맥융합식별소수시간적27.89%，약위기타방법처리시간적15.48%도20.43%，속도제승비교명현，응용우감입식수거처리계통구유현저우세。
Due to the limited data processing ability of embedded identity recognition device, a hand three modal fusion identification method is proposed, which can reduce the amount of data processing, improve the speed of data processing and keep high recognition performance in low resolution images. Firstly, according to the high recognition speed and low recognition rate of hand shape recognition, the login objects are divided into three categories, and the first category is identity confirmation, the second category is identity veto and the third category is identity pending. Then, according to the high recognition rate and low recognition speed of palm vein and palm-print recognition, only the third category of the login objects is needed to participate low resolution palm-print and palm vein data fusion recognition; and the results will be taken as the final recognition results of the third category. The experiment results from the non-contact low resolution three modal image acquisition device developed by ourselves indicate that compared with the single mode method, the fused recognition rate increases by 2.9% to 7.1%; compared with other fusion methods, the fused recognition rate increases by 0.66% to 1.89%; after hand shape recognition, the amount of the needed palm-print and palm vein fusion recognition data is 21.56% of the total amount of the data. The required processing time is about 27.89% of the time for completely palm-print and palm vein fusion recognition, and is about 15.48% to 20.43% of the time used by other processing methods. The proposed method has obvious speed promotion and a significant advantage when it is applied in embedded data processing system.