嵌入式手指多模态特征采集系统对研究基于手指的身份鉴别技术具有重要意义,主要体现在两个方面:一是为理论研究提供可靠的数据,二是为开发实际的身份识别设备奠定技术基础。在前期工作中,已实现了Windows CE下的手指静脉采集系统,但由于Windows CE对高性能硬件系统缺乏完善支持,导致现有单模态系统无法实现功能扩展。面对手指多模态特征采集的实际问题,利用Linux系统开发了嵌入式成像平台,实现了二路视频和一个USB指纹模块的实时成像,从而顺利完成了对手指静脉、指节纹、指纹3个模态的同步采集任务。大量实验结果表明,该采集系统性能稳定可靠,且成像质量能满足理论研究的需求。
Designing a reliable embedded image acquisition system is very important for finger-based biometrics since it can provide reliable multimodal finger image data for finger-based recognition theory research besides making a technical foundation for real identification system development. In previous work, an embedded device with Windows CE operation for finger-vein image acquisition. However, it can not be extended since Windows CE is incapable of supporting hardware with multiple tasks. To deal with the problem faced in muhimodal imaging, a Linux-based embedded system is developed, which can reliably and synchronously output three image modalities : finger-vein, fingerprint and knuckle print in real-time. Lots of experimental results show that the proposed embedded imaging system is stable and reliable in multimodal finger image acquisition, and the imaging quality can meet the demand of theory research.