文章导读
总览 评价 侯亚丽 * , 郝晓莉 , 胡妙春 ( 北京交通大学电子信息工程学院,北京 100044; ) 摘要: 人脸检测在人机交互及未来视频监控等应用中都有着非常重要的作用。传统的人脸检测主要基于RGB图像或视频,极易受到类似人脸如照片、PVC模特等其他物体的
侯亚丽*, 郝晓莉, 胡妙春
(
北京交通大学电子信息工程学院,北京 100044; )
摘要:
人脸检测在人机交互及未来视频监控等应用中都有着非常重要的作用。传统的人脸检测主要基于RGB图像或视频,极易受到类似人脸如照片、PVC模特等其他物体的干扰。多光谱图像中含有更丰富的场景信息,可以利用物体表面不同的反射率进行活体检测,降低系统误检率。本文在窄带多光谱成像系统的基础上,研究了不同的反射率特征提取区域对检测性能的影响,为后续人脸活体检测中多光谱特征提取提供了依据。实验结果表明,不同于传统基于额头区域的特征提取方法,基于整个人脸区域进行特征提取的方法在所测样本集中达到了最佳检测性能。
关键词:
多光谱成像;活体检测;反射率
HOU Yali*, HAO Xiaoli, HU Miaochun
(
School of Electronic and Information Engineering,Beijing Jiaotong University,Beijing 100044; )
Abstract:
Face detection plays an important role in biometric recognition, human-computer interaction, and video surveillance systems. Most traditional human face detection methods are based on RGB images or videos, which can be easily fooled by human-like objects, such as photos, PVC mannequins, etc.. Due to different reflection properties of the object surfaces, techniques based on multispectral images have proved to be effective for liveness detection. In this paper, multispectral feature detection is studied based on a narrow-band multispectral imaging system. In most previous methods, reflectance features from the forehead area are usually used. The experimetal results in this paper show that reflectance features from the entire face area may achieve a better performance.
Tag:
点此返回栏目查看更多>>>参考论文