【作者】 刘晓宁；

【导师】 耿国华；

【作者基本信息】 西北大学 ， 计算机软件与理论， 2003， 硕士

【摘要】 颅像重合技术是基于颅骨的身份认证中一个重要研究方向。应用颅像重合法鉴定无名颅骨身源，可以快速的为侦破提供线索，为司法审判提供科学证据，是法医人类学个人识别的重要技术手段之一。 目前的颅像重合大都采用手工照相技术来实现。本文主要研究计算机辅助颅像重合技术，研究如何在计算机上进行三维颅骨和二维照片的重合鉴定。主要研究内容包括： 1) 鉴定指标定义。在法医人类学和解剖学知识的基础上，分析了颅像重合的鉴定指标，确定了在计算机辅助颅像重合中使用的四项55个鉴定指标。计算机辅助颅像重合就是在这些指标的基础上展开工作的。 2) 研究了如何在计算机上，根据二维照片来调整三维颅骨位置，使人脸和颅骨处于同一位置；讨论了焦距与摄距对颅像重合的影响，以及在计算机这个特定环境下，如何避免这些影响。 3) 研究了在颅像重合特殊要求下对二维照片进行特定预处理，包括图像增强、图像平滑缩放以及图像校准。 4) 面部特征提取。主要研究如何准确提取照片头部轮廓的算法并对其进行实现，采用计算机作为辅助手段进行颅骨和人脸其它特征提取。 5) 颅像配准技术研究。在常规图像配准技术的基础上，将图像配准技术用于颅像重合，给出了具体的基于加权值的颅像配准算法。 在理论研究的同时，我们采用Visual C++6.0以及OpenGL图形包在PC机上设计实现了颅像重合系统PhotoSkull。该系统能对二维照片做特定的预处理工作、准确标定颅骨及照片上的标志点、进行颅像配准。大量试验结果表明，应用本论文中提到的算法进行的颅像配准，有较高的精确度和可靠性。更多还原

【Abstract】 The technology of photographic superimposition is one important direction of personal identification which based on skull. Using the photographic superimposition to identify a body without a name, it can provide some clues for solving a case and scientific proof for judicial review. It is an important personal identification way in forensic anthropology.Most of present superimposition are realized by taking photos. This thesis mainly research on computer-aided photographic superimposition Technique, study how to make a 3D skull and a 2D photo be superimposition in computer. This paper mainly include:1) Definition of identification features. Based on forensic anthropology and anatomy, we analyze features used in superimposition, and then select four items, fifty-five features, which be based on by computer-aided photographic superimposition.2) Study how to adjust the skull according the photo in computer, and how to avoid the influences from the distance of focus and photography.3) To satisfy the specific requirements of photographic superimposition, we should preprocess the 2D-Photo, including image boosting, image scaling without loss and image adjusting.4) Extract the features on face and skull. Study how to get contour of head and other characters on face and skull.5) Study skull-photo registration. Based on traditional image registration, we apply it to photographic superimposition, given the specific detail steps.Besides these theory researches, we also built a photographic superimposition system, which named PhotoSkull, using Visual C++ and OpenGL on the PC platform, which can preprocess images, sign the feature points on skull and photo and then register the skull and photo. A lot of tests tell that the skull-photo a registration using the algorithms in this thesis can get high precision and reliability.更多还原