【作者】 郭德军；

【导师】 宋蛰存；

【作者基本信息】 东北林业大学 ， 控制理论与控制工程， 2005， 硕士

【摘要】 皮肤表面是人体生理健康状况的一面镜子,如何客观、定量地评价面部皮肤当前的状态一直是医学界和化妆品界关注的问题。本文在现有的理论和文献基础上,研究了一种基于图像纹理统计特征分析的面部皮肤状态检测系统。 本文首先介绍了客观定量评价面部皮肤的状态在医学和皮肤美容中的重要意义。总结了目前国内外的皮肤表面状态检测技术,深入研究了各种纹理的统计特征分析方法,包括空间灰度共生矩阵法、中心矩方法、灰度差分统计分析方法、游程长度法,定性地分析了这几种统计方法中各个特征的实际纹理意义,提出了采用空间灰度共生矩阵法提取皮肤纹理的特征。 然后,本文通过对现有图像预处理方法的实验分析,提出了适合对皮肤图像进行预处理的方法。接着,介绍了神经网络模式识别技术及其目前的研究发展现状,在研究了传统的BP网络算法的基础上,采用了一种改进BP网络算法作为人工神经网络模式识别的分类识别算法。在MATLAB环境下编写了皮肤图像的预处理和特征值提取程序,设计了基于CCD摄像机,图像采集卡和计算机的面部皮肤图像采集系统。 最后在实验部分验证了采用空间灰度共生矩阵法提取皮肤纹理特征的准确性,并利用MATLAB中的神经网络工具箱对BP网络进行了训练和识别。实验的结果证明了本文所提出的基于图像纹理统计特征分析的面部皮肤状态检测系统方案的可行性和理论分析的正确性。更多还原

【Abstract】 Skin surface is a mirror of human health condition, and that how to evaluate current condition of facial skin objectively and quantitatively has been the problem regarded by both medical and cosmetic group. In this paper, depending on available theory and literature, it studied a kind of measuring system of facial skin condition based on statistical feature analysis of image texture.Firstly, it introduced the important significance of evaluating facial skin texture objectively and quantitatively in the field of medicine and skin hairdressing, sumed up the measuring technology of skin surface condition in home and abroad at present, and researched into diversified means of statistical feature analysis of image texture, including spatial gray level co-occurrence matrix means, center square means, gray level difference statistical analysis means, traveling length means. Then it analyzed the actual physical significance of every feature in these statistical means qualitatively, and put forward to extract the features of skin texture with spatial gray level co-occurrence matrix means.Secondly, it put forward a suitable way to process the skin image by testing analysis to available image pre-processing means. Then it introduced the neural network pattern recognizing technology and its current study development status. An improved BP network arithmetic was adopted as artificial neural net pattern recognizing arithmetic after reseaching the traditional arithmetic of BP network. It compiled the programs to process skin image and to extract the feature with MATLAB soft. It designed a collection system of facial skin image based on CCD camera, image collection card and computer.In the end, it validated the validity with spatial gray level co-occurrence matrix means extracting the skin texture feature in the experiment part, and made use of the neural network tools of MATLAB soft for BP network to train and recognize. The experiment results prove that the measuring system project of facial skin condition based on statistical feature analysis of image is feasible and the academic analysis is correct.更多还原