基于近红外眼周图像有用特征的人类身份识别研究

The research of human identity recognition based on useful features of near-nfrared periocular images

针对现存身份识别算法很少有效利用眼周图像的问题,本文执行了两组实验探究人类如何分析眼周图像。首先,选择和预处理用于实验的眼睛图像;然后,以对象随机配对的方式创建不同对象的查询,发现志愿者能对92%的查询正确确定两幅图像的关系;最后,在形成不同对象对时考虑了多个因素,从具有相同年龄和种族,类似的眼睛颜色、眼妆、睫毛长度和眼睛遮挡的对象对形成查询,并且限制志愿者观察查询对的时间。实验取得的正确验证率为79%,分析结果表明,在现有身份识别系统中合并使用眼周识别算法可以更加准确有效地进行人类身份识别。

To solve the problem of the periocular images not effectively exploited in current human identification algorithm, this paper conducted two experiments to determine how humans analyze periocular images. Firstly,the eye images for our experiment are selected and pre-processed. Then subjects were paired randomly to create different-subject queries and show that the volunteers correctly determined the relationship between the two images in92% of the queries. Finally, multiple factors are considered in forming different-subject pairs; queries were formed from pairs of subjects with the same gender and race, and with similar eye color, makeup, eyelash length, and eye occlusion. In addition, the amount of time volunteers could view a query pair is limited. In this experiment, the correct verification rate was 79%. The analysis shows that the human identification is effectively when combing the human identification system and the periocular identification algorithm.