关于PSO人脸姿态估计.PDFVIP

第 14 卷 第 2 期 太赫兹科学与电子信息学报 Vo1．14，No．2 2016 年 4 月 Journal of Terahertz Science and Electronic Information Technology Apr．，2016 文章编号 ：2095-4980(2016)02-0282-06 基于 PSO 的人脸姿态估计 雷 欢 a,b ，陈 虎 a,b ，吴志红 a,b* ( 四川大学 a.计算机学院； b.视觉合成图形图像技术国防重点学科实验室，四川 成都 610065) 摘 要 ：提出了一种人脸姿态估计的新方法。根据三维扫描预先生成一个通用的模型，基于 三维形态模型采用粒子群优化算法对人脸姿态进行初步估计，由特征点检测确定姿态大致范围， 再在初步估计的结果上进行修正迭代，从而对人脸姿态进行精确估计。实验表明，该方法在简化 数学运算的基础上，具有较好的估计效果，平衡了计算复杂度与结果精确度。 关键词 ：人脸姿态估计；三维形态模型；粒子群优化；迭代 中图分类号 ：TN911.73 文献标识码 ：A doi ：10.11805/TKYDA201602.0282 Face pose estimation based on PSO a,b a,b a,b * LEI Huan ，CHEN Hu ，WU Zhihong (a.College of Computer；b.National Key Laboratory of Fundamental Science on Synthetic Vision，Sichuan University，Chengdu Sichuan 610065，China) Abstract：The face pose estimation based on Particle Swarm Optimization(P SO) is proposed. Firstly, a common model is generated by 3-D scanning. Then, PSO algorithm is used to estimate the face pose preliminarily based on the 3-D morphable model to acquire the pose range approximately with a detection of feature points. Finally, the rectification and iteration are applied with the preliminary result to estimate the face pose correctly. Experimental results show that the proposed method can obtain satisfying result by simplifying the math arithmetic, which keeps a balance between the complexity and the precision, and has an good effect on further studying of face identification. Key words：face pose estimation；3 -D morphable model；Par