基于超像素分割的IPPG活体皮肤检测  

IPPG Alive-Skin Detection Based on Superpixel Segmentation

作　　者：孔令琴[1] 吴育恒 赵跃进[1] 董立泉[1] 刘明[1] 惠梅[1] Kong Lingqin;Wu Yuheng;Zhao Yuejin;Dong Liquan;Liu Ming;Hui Mei(Beijing Key Laboratory for Precision Optoelectronic Measurement Instrument and Technology,School of Optics and Photonics,Beijing Institute of Technology,Beijing 100081 China)

机构地区：[1]北京理工大学光电学院精密光电测试仪器与技术北京市重点实验室,北京100081

出　　处：《光学学报》2020年第13期35-42,共8页Acta Optica Sinica

基　　金：国家自然科学基金(61705010,61935001,11774031)。

摘　　要：针对现有活体皮肤检测方法精度不高、实时性较差的问题,提出一种基于超像素分割的成像式光电容积描记(IPPG)活体皮肤检测(SPASD)算法.利用零参数简单线性迭代聚类算法将图像分割为多个超像素子块;然后通过IPPG技术并行提取各子块中的脉搏波信号;最后利用支持向量机对提取到的信号进行训练分类,进而实现活体皮肤的实时检测.实验结果表明,SPASD算法可以有效提高活体皮肤的检测精度和实时性,其检测精度达92.02%.所提方法在人脸防骗、非接触生理信号检测、面部表情识别等领域具有应用前景.The existing alive-skin detection methods exhibit low accuracy and poor real-time performance.Therefore,an image photoplethysmography(IPPG)alive-skin detection(SPASD)algorithm is proposed based on superpixel segmentation in this study.An image is segmented into multiple superpixel sub-blocks using the simple linear iterative clustering zero-parameter algorithm;subsequently,the IPPG technology is used to extract pulse signals from each sub-block in parallel.Finally,a support vector machine is used to train and classify the extracted signals for achieving real-time alive-skin detection.The experimental results demonstrate that the SPASD algorithm can effectively improve the alive-skin detection accuracy(92.02%)and real-time performance.The proposed method can be applied in face anti-fraud,non-contact physiological signal detection,facial expression recognition,and other fields.

关 键 词：图像处理 模式识别 成像系统 成像式光电容积描记 活体皮肤检测 超像素 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]