基于三维数据的人脸表情识别-模式识别与智能系统专业论文.docx
文本预览下载声明
摘要迄今为止,表情识别领域的绝大多数工作都是以二维静态或序列图像为
摘要
迄今为止,表情识别领域的绝大多数工作都是以二维静态或序列图像为 处理对象的。这些方法给出的表情识别率~般都较低,而且能识别的表情的 种类也不多,尤其是对微小的表情变化难以识别,成为表情识别研究中的一 大难点。另外,大多数已有的人脸表情识别算法的适应性较差,不能完成对 陌生人面部表情的识别工作,在光照条件不同以及待识别人的头部姿态存在 变化的情况下,也不能取得满意的识别结果。近年来,随着三维成像技术的 发展,三维图像的获取越来越容易。和二维图像相比,三维图像可以提供信 息量更大、鲁棒性更强并且与光照和人的头部姿态等因素无关的三维信息。 显然,将人脸三维图像用于人脸表情识别,可以期望得到更好的结果。
现在进行的大多数的人脸表情识别研究都利用的是二维的人脸图像。三 维成像技术的发展使得三维图像的获得变得容易而且被用在越来越多的地 方。为了充分利用三维图像,在前人的二维图像研究的基础上,本文提出了 一种新的基于三维数据的人脸表情识别方法。此方法根据三维成像仪的特性, 利用得到的三维数据,并结合原始的二维图像进行特征点识别,然后综合分 析特征点的性质完成人脸表情识别。实验表明,本方法能够对六种人脸表情 进行识别,有较高的识别率。
关键词:表情识别、三维数据、特征点
AbstractTi|I
Abstract
Ti|I now,a Iot of methods have been presented to solve the problem of facial expression recognition.These methods usually use 2D static images or image sequences to perform the recognition tasks.It is unfortunate thal most of them have only Iow recognition rate.They can only recognize a smalI set of expressions Especially,most of them cannot recognize subtle expression changes.In addition.the adaptability of many currenl algorithms is low,they can hardly recognize the expressions of strangers.
and in the cases of differenl lighting conditions and(or)different head poses
satisfactory results are barely obtained.Recently,with the development of 3D imaging techniques,it is easier to get 3D images now Compared with 2D Image.3D image can provide 3D lnformation that has more capacily and is more robust,independent of lighting condition and head pose.Thus, it is expected thal better results can be obtained by utilizing 3D faciaI
Images.
In order to make fulI use of 3D images.we preposed a new 3D data—based expression recognition method.The method uses both original 3D data and 2D color image provided by a 3D imaging instrumenl to extract feature
points and then to recognize expressions aCCOrding to these feature points The experimentaI results show that six facial expressions can be recognized in a high recognition rate.
Key words:facial expression recognition,3D image,feature point
致谢此
显示全部