标准正面人脸图像的识别

本论文选用人脸上27个特殊点作为人脸基本特征。以人脸几何结构为基础,结合有脸识别的心理特性,提出新颖、简便、高精度的“寻找存在”法,使提取特征点的速度、精度得到大大的提高,在详细分析这27个特列点的统计特性后,选择了其中信息量丰富的15个点间距及间距比构成一组向量代替人脸描述,用加权欧氏距离作为特征向量间相似性测试,在两类实验中,识别率高达100％和98％。

RECOGNIZING A TYPICAL HUMAN FACE FROM THE FRONT SIDE

In order to recognize a person's face fast and perfectly, 27 points on the face are extracted as feature points automatically by using "Searching the existence" method. The distance between these points and the ratios between these distances serve as parameters to discribe the facial features, from which 15 points are selected by analysing their statistical characters. Recognition accuracy reachs 100% and 98% respectively in two different experiments according to the recognition criterion, the weighted Euclidean distance. The system takes 1 minute and 50 seconds from sending a face image to finishing the recognition on an IBM PC/AT.