基于面象特征的中医体质自动辨识系统研究

发布时间：2018-01-12 15:33

本文关键词：基于面象特征的中医体质自动辨识系统研究　出处：《北京工业大学》2016年硕士论文　论文类型：学位论文

【摘要】：中医学认为人体是一个有机整体,需要根据个体体质进行养生保健和疾病防治。所以,面部形态、神色的变化与身体体质有着某种程度上的联系。目前的面诊研究主要集中于面色与脏腑病症或其他具体病症之间的关系研究。本论文通过拍摄受检者的面部图像,利用数字图像处理、模式识别等技术初步完成了中医面诊的自动分析研究,并在此基础上,完成了一个体质辨识的辅助系统。该系统为体质辨识提供更多的理论依据,对中医四诊客观化的发展和中医体质辨识有重要的现实意义。具体来说,本论文主要完成以下几方面的工作:1.完成了对采集到的人体面部图像的预处理工作。在面象采集过程中,由于人体姿势或拍摄角度等因素的影响,我们采集到的人体面部图像不统一。为了排除其他不稳定因素的干扰,减小误差,得到统一的人体面部图像,有必要对采集的人体面部图像进行预处理工作。本论文利用基于YCb Cr颜色空间的椭圆模型进行肤色检测,并对检测结果进行去噪处理,进而利用基于面部矩形特征的方法进行人脸定位,最终得到图像中人脸的具体位置。2.提取与中医体质类型相关的面象颜色特征和纹理特征。对面部额头、鼻子、脸颊进行子图定位,然后对其颜色特征和纹理特征进行提取,将人脸上的特征转化为数字形式的向量。在颜色特征提取时,将图像按非线性转换公式由RGB颜色空间转换到HSV颜色空间,然后提取面部子图的H、S、V三个分量作为面象的颜色特征;对于纹理特征,利用灰度共生矩阵的四个分量表征纹理特征。3.首次对面象特征与中医体质的关系进行了客观化的研究。在图像预处理和特征提取的工作基础上,采用SVM对面象特征进行分类学习,分别使用网格遍历法和粒子群法对SVM中核参数?和惩罚因子C进行寻优,并结合交叉验证法得到最优参数,进而对基于面象特征的中医体质进行分类研究,探索了面象特征与中医体质的客观关系。4.开发了基于面象特征的中医体质自动辨识系统。在以上研究的基础上,搭建全生命周期健康管理平台,设计开发面象自动分析模块,对中医体质进行自动辨识,对中医全面健康体检提供新的依据,也为基于面象特征的中医体质辨识提供了一个新的思路。
[Abstract]:Chinese medicine believes that the human body is an organic whole, according to the individual health care and health care and disease prevention and treatment. Therefore, facial morphology. There is a certain relationship between the changes of facial appearance and physical fitness. The current study of facial examination mainly focuses on the relationship between facial color and viscera disorders or other specific diseases. This paper takes the face images of the subjects. . Using digital image processing, pattern recognition and other technologies, we have preliminarily completed the automatic analysis of traditional Chinese medicine surface diagnosis, and on this basis. The system provides more theoretical basis for physique identification, and has important practical significance for the development of TCM four diagnosis objectification and TCM physique identification. The main work of this thesis is as follows: 1. The preprocessing of the collected human face image is completed. In the process of face image acquisition, due to the human posture or shooting angle and other factors. In order to eliminate the interference of other unstable factors and reduce the error, we can get the unified human face image. It is necessary to preprocess the collected human face image. In this paper, we use the elliptical model based on YCb Cr color space to detect the skin color, and de-noising the detection results. Then the face location based on the face rectangular feature is used to get the specific position of the face in the image. The color and texture features of the face image are extracted which are related to the physique type of traditional Chinese medicine. The nose and cheek are located by subgraph, then the color feature and texture feature are extracted, and the feature of human face is transformed into a vector in the form of number. When the color feature is extracted, the color feature is extracted. The image is transformed from RGB color space to HSV color space according to the nonlinear transformation formula, and then three components of the face image are extracted as the color features of the face image. For texture features. Using four components of gray level co-occurrence matrix to characterize texture feature. 3. The relationship between the first image feature and TCM physique is studied objectively. Based on the work of image preprocessing and feature extraction. The SVM image features are used for classification and the mesh ergodic method and particle swarm optimization method are used to study the kernel parameters in SVM. And penalty factor C to optimize, and combined with cross-validation method to obtain the optimal parameters, and then based on the features of traditional Chinese medicine physique classification research. Explore the objective relationship between facial features and TCM physique. 4. Develop an automatic identification system of TCM physique based on facet features. On the basis of the above research, build the whole life cycle health management platform. The design and development of automatic surface image analysis module for automatic identification of TCM physique provides a new basis for comprehensive physical examination of TCM and also provides a new idea for TCM physique identification based on facial features.
【学位授予单位】：北京工业大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：R241;TP391.41

【相似文献】