【作者】 卢曼慧；

【导师】 苑玮琦；

【作者基本信息】 沈阳工业大学 ， 计算机应用技术， 2007， 硕士

【摘要】 随着计算机技术的迅速发展，信息安全问题逐渐受到人们的重视，于是为了满足信息时代的安全要求，产生了生物特征识别技术。目前常用的生物特征识别技术主要有人脸识别、指纹识别、虹膜识别、人耳识别等，其中人脸识别以其采集方便、用户接受程度高、实用性强的优势受到人们的关注，但是，由于年龄、表情、化妆和胡须等变化的影响，使得人脸的特征很不稳定，故人脸识别具有一定的局限性。相对于人脸图像，人耳图像不受表情、化妆等的影响，而且也具有采集方便、用户接受程度高的优点，因此将人耳识别与人脸识别相结合可以真正做到“非打扰识别”。根据人耳特殊的生理特征结构和生理位置，本文提出了一种将人脸特征和人耳特征相结合的基于侧脸和耳廓组合特征的个人身份鉴别方法，用此方法来辅助人脸识别，提高识别率。基于侧脸和耳廓组合特征的个人身份鉴别方法分别提取了侧脸特征和耳廓特征，并将其相结合进行识别。首先对图像进行预处理去掉图像中的背景信息和噪声，然后用轮廓跟踪的方法在侧脸图像中提取侧脸轮廓，根据曲率提取眼睛、鼻子、嘴三个特征点，以其两两之间的垂直和水平距离作为侧脸特征。在提取完侧脸特征之后，提出一种基于轮廓跟踪的人耳定位算法，用该算法进行人耳定位，提取外耳轮廓，然后提取外耳轮廓的高度和宽度特征作为耳廓特征。最后将侧脸特征和耳廓特征相结合形成组合特征向量，采用根据相似性阈值和最小距离原则的聚类方法对其进行分类识别。应用实验室自建的侧脸图像库对本文提出的方法进行验证，实验结果表明该方法是可行的，可以有效地辅助人脸识别。而且该方法的人耳定位部分提取出的外耳轮廓及定位出的人耳图像还为进一步提取耳廓及内耳特征打下良好的基础。更多还原

【Abstract】 With the development of computer technology, people attach importance to the information security. In order to meet demands of information age, biometrics appears. At present, biometrics consists of face recognition, fingerprint recognition, iris recognition, ear recognition etc. Due to easy getting, nicety acceptance and great practicability, face recognition is attached great importance. However, face recognition is restricted because of the change of age, appearance, makeup and beard. Compared to face image, ear image can’t be affected by change of appearance, makeup and keep the advantage of easy getting, thus combining ear recognition and face recognition can be no disturbance recognition. According to the special construct and position of ear, this paper proposes a personal identification based on the combinatory features of profile and outer ear contour, which can be the remarkable enhancement of the face recognition performance.The method based on combinatory features of profile and outer ear contour extracts the features of profile and outer ear contour firstly, and then, combines them each other. First, take out background and yawp of images. Then, the profile of side face image is extracted by edge tracking. According to the curvature of profile, three feature points whose vertical distance and horizontal distance are taken as feature are extracted. After that, utilize an ear localization algorithm based on edge tracking to extract the outer ear contour. Combine the height and the width of the outer ear contour with the profile feature. Finally, make eigenvector classification with clustering method based on close threshold and minimum distance.Utilize the method in the side face gallery. The experiment result shows that this method is feasible and can assistant face recognition effectively. At the same time, the outer ear contour and ear image extracted can be used to extract the other features further.更多还原