摘要
针对传统人脸检测系统的不足,提出了一种基于DSP的人脸检测系统的设计方案。分析了该系统实现的基本原理,通过对各种检测方法的比较,选择了肤色检测法为基本手段。利用人脸样本库建立了人脸肤色模型,讨论了形态学滤波的优势,随后提出了利用人脸的结构特征,进行区域确认的方法。最终,从硬件和软件这2个方面对该系统进行了总体规划,并阐述了对DSP程序进行优化的相关问题。实验结果表明,该系统能较为准确地检测人脸,并满足了实时性的要求,具有一定的实用价值。
Because of the limitations of traditional face detecting system, a new system based on DSP is put forward. Principles of the new system are analyzed. By the comparison of many detecting methods, the skin - color detecting method is chosen as the basic means. A human face skin - color model is built by the human face database. Subsequently, predominance of morphological filtering is discussed and a face region affirming method based on face structure is presented. Finally, the framework of the system＇ s hardware and software are provided. Relative problems of DSP program optimization are discussed. Experimental results show the system meets the real -time demand and is practicable, which can the face correctly.
出处
《武汉理工大学学报（信息与管理工程版）》
CAS
2007年第10期1-4,9,共5页
Journal of Wuhan University of Technology：Information & Management Engineering
关键词
人脸检测
DSP
肤色模型
形态学滤波
face detection
DSP
skin - color model
morphological filtering
