Three-dimensional facial feature points matching based on K-means clustering of relative angle context distribution and support vector machine

doi:10.11834/jig.20110501

Views : 0 下载量: 223 CSCD: 0

PDF
Export
Share
Collection
Album

Three-dimensional facial feature points matching based on K-means clustering of relative angle context distribution and support vector machine
Vol. 16, Issue 5, Pages: 850-856(2011)
Published Online：23 May 2011，

Published：2011
DOI： 10.11834/jig.20110501
稿件说明：

移动端阅览

DOI：

Three-dimensional facial feature points matching based on K-means clustering of relative angle context distribution and support vector machine[J]. Journal of Image and Graphics, 2011, 16(5): 850-856. DOI： 10.11834/jig.20110501.

摘要

人脸特征点自动定位及对应点匹配是计算机视觉和模式识别领域一个非常热门的研究方向，应用领域包括图像配准、对象识别与跟踪、3维重建、立体匹配等。通过相对角直方图分布和K均值聚类确定脸部特征点的聚类点集，再利用几何信息提取聚类点集的特征，进而采用支持向量机分类最终从点集中分离出39个脸部特征点。实验结果表明，此混合提取方法比单纯使用RAC得到了更好的匹配准确率，在给定的距离阈值范围内，50%的特征点定位准确率达到了100%。

Abstract

Feature points searching or point correspondence matching is a challenge in computer vision and pattern recognition

which is very important perquisite for many 2D/3D applications such as image registration，object recognition and statistical model construction. In this paper

we propose an algorithm for facial feature points matching among 3D point cloud models. Specifically

the surface points are clustered based on relative angle context (RAC) features

and then the geometric features of the clustered points are extracted. Afterwards

supported Vector Machine based classification is employed for final accurate correspondence location. The experimental results demonstrate that our algorithm achieves better performance than RAC algorithm proposed.Within the confines of a given distance threshold

the accuracy rates of 50% feature points have even reached to 100%.