曲面迭代的混沌特性及其在人脸识别中的应用

Image Chaotic Characteristics and Application in Face Recognition

为探讨混沌理论在图像应用中的更多可能性,减小图像识别过程中因局部动态变化等因素对识别率的影响,提出一种基于混沌迭代的图像特征构造方法.首先利用图像与辅助函数构造离散动力系统,在二维空间中使用Euler法进行迭代,得到近似吸引子作为图像的特征点阵;然后对该特征点阵进行Radon变换,将其投影到一维空间中,通过计算相关系数等方法对人脸图像进行识别;给出一种灰度自适应方法,在图像识别过程中通过调整灰度对比度来提高吸引子生成质量.实验结果表明,利用该单一特征的初步识别方法在Yalefaces人脸数据库中的识别率为70.91%;改进灰度自适应等方法后识别准确率达到87.33%,可得到较稳定的识别效果;另外,通过绘制出一些不同类图像的吸引子,说明图像不同其吸引子形状也不同.

To explore the widely possibilities of chaos theory in image application, this paper investigates a new image feature construction method based on chaos iteration to reduce the impact of local dynamic changes in the image recognition. The discrete dynamical system is constructed by the image and the auxil-iary function. And the approximate attractors which are considered as the image feature lattice got from the discrete dynamical system through Euler iteration. The paper achieves the face recognition by calculating the correlation coefficient of the feature lattice which is projected into the one-dimensional space according to the Radon transform. A gray-adaptive method is proposed to improve the quality of the attractors by adjust-ing the gray contrast. Experiments show that with the proposed improved gray adaptive method, the identi-fication rate of single feature in the Yalefaces rises from 70.91% to 87.33%. In addition, the experiment in-dicates that the shapes of the three-dimensional attractors vary with the image sequence.