查询结果:   张雪峰,赵莉.视觉单词优化仿射尺度不变特征变换的视频人脸识别[J].计算机应用与软件,2015,32(7):223 - 227.
中文标题
视觉单词优化仿射尺度不变特征变换的视频人脸识别
发表栏目
图像处理与应用
摘要点击数
794
英文标题
VIDEO FACE RECOGNITION WITH AFFINE SCALE INVARIANT  FEATURE TRANSFORM OPTIMISED BY VISUAL WORDS
作 者
张雪峰 赵莉 Zhang Xuefeng Zhao Li
作者单位
信阳农林学院计算机科学系 河南 信阳 464000     
英文单位
Department of Computer Science, Xinyang College of Agriculture and Forestry, Xinyang 464000, Henan, China     
关键词
视频人脸识别 尺度不变特征变换 仿射SIFT 视觉单词 巴氏距离
Keywords
Video face recognition Scale invariant feature transform Affine-SIFT Visual words Bhattacharyya distance
基金项目
作者资料
张雪峰,讲师,主研领域:模式识别,图像处理。赵莉,讲师。 。
文章摘要
针对视频人脸识别中由于光照、表情、姿态等变化而影响识别性能的问题,提出一种基于视觉单词优化仿射尺度不变特征变换的视频人脸识别算法。首先从兴趣点提取仿射尺度不变特征变换的图像描述符,将其作为人脸图像表示法;然后,由高斯差分检测,使用视觉单词的索引取代这些描述符;最后,计算视觉单词之间的巴氏距离,并利用最大相似性原则完成识别。在两大通用视频人脸数据库Honda及MoBo上的实验验证了该算法的有效性。实验结果表明,相比其他几种较为先进的视频人脸识别算法,该算法明显提高了识别率,并且大大降低了计算复杂度,有望应用于实时视频人脸识别系统。
Abstract
The recognition performance is impacted by the variation of illustration, expression and gestures in video face recognition, for this problem, we propose a video face recognition algorithm which is based on optimising affine scale invariant feature transform (SIFT) with visual words. Firstly, it extracts from interest points the image descriptors of SIFT and uses them as the image representation method. Then, with Gaussians difference detection, it uses the indexes of visual words to replace these descriptors. Finally, it calculates Bhattacharyya distances between the visual words and employs maximum likelihood principle to complete face recognition. The effectiveness of the proposed algorithm is verified by experiments on Honda and MoBo, experimental results show that the proposed algorithm remarkably improves recognition rate and decreases computation complexity compared with several other advanced video face recognition algorithms, and could be expected to apply to real-time video face recognition systems.
下载PDF全文