【作者】 孙晓玲；

【导师】 侯德文；

【作者基本信息】 山东师范大学 ， 计算机软件与理论， 2008， 硕士

【摘要】 生物识别技术是利用人自身具备的生物特征进行身份鉴别的技术。与其他的生物认证技术相比,人脸识别具有更直接、友好、方便的优点。因此,人脸识别技术有着广泛的应用前景和迫切的现实需要,是当前模式识别领域最热门的研究方向之一。多年来,人们对人脸识别技术中的许多问题都进行了深入地研究,并且已经研制出了不少有效的算法。然而,由于不同人脸具有的内在相似性,而同一人脸的不同图像则常常因为一些因素的影响而表现出巨大的差异性,因此现有的人脸识别技术仍然无法满足实际应用的需要。一般情况下,一个完整的人脸识别系统由人脸检测、特征提取、人脸识别等几部分组成。其中,特征提取的好坏将直接影响到识别的效果。而在脸部的所有特征点中,眼睛是最重要的器官,眼睛定位的准确与否将影响到特征提取的效果。因此人眼的快速准确定位被学者视为提高人脸识别准确率和人脸识别走向实用化的关键和瓶颈所在。本文以实现眼睛的准确定位为目标,探讨了基于脸部的肤色、灰度和几何特性的眼睛定位方法。整个过程可以分为两个阶段,第一阶段研究了从图像中检测定位人脸的方法;第二阶段又研究了在脸部区域进行眼睛定位的方法。论文的研究工作主要包括以下内容:第一,分别研究了适用于两种不同情况的脸部区域检测算法,算法根据其不同的应用特点进行了不同的处理。对于特定场合背景较简单的情况,采用投影法对图像进行了处理,然后通过分析其投影曲线的特征,进行脸部的粗定位。这种方法简单易实现,且计算复杂度小,非常适合处理一些实时的任务。而对于复杂背景的情况,则根据肤色对尺寸、姿态、表情不敏感的特性,采用基于肤色信息的方法来进行人脸检测。首先对不同的色彩空间进行分析比较,并选取YCbCr空间进行研究,然后通过大量肤色样本建立的高斯模型进行肤色分割,最后通过脸部的几何特性对分割出来的肤色区域进行优化处理。实验证明,文中针对两种不同情况采用的方法,在各自的应用场合中都取得了较好的效果。第二,在检测到的脸部区域内进行眼睛定位阶段,本文提出了一种有效的眼睛精确定位方法。由于眼睛区域具有复杂的灰度变化特征,所以根据其灰度信息通常可以实现准确的定位。本文的方法是先对图像进行形态学处理从而粗定位到候选的眼睛点;而眼睛的精确定位过程采用的是一个复合的眼睛检测器,该检测器由两个不同的方差滤波器级联构成,通过构造的检测器和先验知识的验证,可以判定出眼睛的精确位置。实验证明,本文采用的眼睛定位方法容易实现,且具有较好的鲁棒性。更多还原

【Abstract】 Biometric identification technology is a kind of technology using individual biometric characteristics to verify identity. Compared with other biometrics technology, face recognition is more direct, friend and convenient. So the technology of face recognition is one of the hottest research directions for its wide application and urgent demand. Face recognition has been intensively investigated by researchers and many useful algorithms have been developed. Since faces of different subjects are often similar, while face images from the same person often differ quite significantly due to the impact of certain factors, current face recognition systems cannot meet the requirements of many practical applications.Usually a complete face recognition system includes face detection, feature extraction and face recognition. The quality of feature extraction will be affected by the recognition effects directly. The eye is the most important organ of face. As the accuracy of eye detection will affect the quality of feature extraction, rapid and precise eye detection is considered the most crucial work to be dealt with.In this paper, in order to locate the eye exactly, the methods based on the facial color, grayness and geometric characteristics are studied. The whole process can be divided into two phases, that is, face detection and eye location in the face region. The main contributions are summarized as follows:Firstly, in this paper two algorithms of face detection to different situations are studied, and they do corresponding treatments according to the different applications. For the specific occasions of simple background, gray projection is adopted to process the images. Face can be located roughly through analyzing the characteristics of the projection curve. This method is easy to implement and costs small complexity, so it can be adapted to deal with some real time tasks. For complex background, because the skin-color is not sensitive to size, pose and expression, the method based on skin-color is used to detect the face. By comparing the different color space, YCbCr color space is adopted. Then Gaussian model is established to segment the image and face region can be validated according to geometric characteristics. The result of experiment indicates that these two algorithms are all effective. Then an effective eye location algorithm in the face region is brought forward. Because of the complex gray of the eye region, these characteristics can be used to locate the eyes exactly. Eye candidates can be found through morphology operation; Eye detector is designed, and it is structured with two different variance filters. The eyes are accurately located with the detector and prior knowledge. The experimental results prove that the algorithm is easy, efficient and robust.更多还原