基于计算机视觉的未识别民族面部特征研究

基于计算机视觉通过特征点定位方法提取图像中的人脸特征点,利用其构建了眉间距、瞳孔间距、眼睛宽度等多项指标来刻画民族人脸特征,并分析了穿青人、南京人、蔡家人和革家人等未识别民族的特征,且与汉族、朝鲜族、藏族、蒙古族、壮族和维族等民族进行了相似性度量。实验表明未识别民族的面部特征分布存在一定规律,但受性别影响,不同民族面部之间的相似性也不尽相同。研究结果不仅证实了利用计算机视觉技术对多民族人脸特征进行分析的可行性,也为民族学人类学的相关研究提供了一种新途径。

The Study for Facial Features of Unrecognized Ethnic Groups Based on Computer Vision

The facial feature points are first extracted from images using landmark method in this paper. Then, the facial features, such as brow distance, pupil distance, eye wide, ect. are constructed for describing facial characteristics. Last, unrecognized ethnic groups such as Chuanqing, Nanjing, Caijia and Gejia are analyzed, and the similarity measurement is also conducted by clustering method among multi-ethnic groups including Chuanqing, Nanjing, Caijia, Gejia, Han, Korean, Tibetan, Mongolian, Zhuang and Uygur. The experiments show that there are regularities in the distribution of unrecognized ethnic groups. Due to the effect of gender, the similarities are also different among multi-ethnic groups. The results not only confirm the effectiveness for facial feature analysis using computer technology, but also provide a new approach for anthropology research.