Face recognition under varying pose and lighting--towards automatic personal identification for surveillance systems

不同姿势和光照下的人脸识别——面向监控系统的自动个人识别

• 批准号: DP0451091
• 负责人: Prof Yongsheng Gao
• 金额: $ 11.49万
• 依托单位: Griffith University
• 依托单位国家: 澳大利亚
• 项目类别: Discovery Projects
• 财政年份: 2004
• 资助国家: 澳大利亚
• 起止时间: 2004-01-01 至 2008-12-31
• 项目状态: 已结题
• 来源: https://dataportal.arc.gov.au/RGS/Web/Grants/DP0451091
• 关键词: Face; recognition; under; varying; pose

One of the key remaining problems in computerized human face recognition is the need to handle the variability in appearance due to changes in pose. This proposed research targets at identifying a person with a face image in a pose different from the example view by using a novel texture analysis and synthesis technique. This technique makes use of facial textures at different viewing directions and can recover appropriate textures for virtual views in arbitrary poses. The successfulness of the proposed research would make a technical breakthrough towards solving the major remaining problem in face recognition.

计算机化人脸识别的关键问题之一是需要处理由于姿势变化而导致的外观变化。这项研究旨在通过一种新的纹理分析和合成技术来识别具有不同于示例视图的姿势的人脸图像的人。该技术利用了不同观察方向的人脸纹理，可以恢复任意姿势下的虚拟视点的适当纹理。拟议研究的成功将在解决人脸识别方面的主要遗留问题方面取得技术突破。