面向智能终端的眼镜虚拟试戴系统

发布时间：2018-03-26 11:35

本文选题：眼镜虚拟试戴系统　切入点：人眼检测　出处：《南京理工大学》2017年硕士论文

【摘要】：现如今计算机和互联网技术迅猛发展,各种软件和商业模式应运而生。同时,由于硬件技术的不断提高,智能终端的性能有了飞跃性的改变,其在人们日常生活中扮演着重要的角色,用户数量越来越多。随着移动互联网的兴起,"互联网+"成为一种新的经济态势,将传统产业和互联网结合起来成为一种更好的发展方式,越来越多的系统和软件被开发出来应用在传统行业中,焕发出了行业新的创新活力。眼镜主要分为隐形眼镜和框架眼镜两种,而隐形眼镜中美瞳眼镜因其美颜效果受到越来越多年轻人喜爱。对于眼镜行业来说,目前的主要问题是用户选择美瞳或者框架眼镜样式都需要多次重复真实试戴,不友好且不卫生,针对这一问题本文提出开发一个面向智能终端的眼镜虚拟试戴系统软件,实现美瞳试戴、框架眼镜试戴、资讯和更多功能,以提高购买眼镜的用户体验。其中,智能终端性能有了巨大的进步,但运算处理能力无论如何都无法与计算机相提并论,所以如何在智能终端内存较低、计算速度较慢的条件下实现人眼和虹膜特征的精确定位以及不同图像无缝融合是本系统开发完成的关键。本文工作和主要内容是按技术模块进行划分的,分别是特征检测、图像融合和Android开发。首先特征检测模块,人眼特征采用了 AdaBoost算法对输入图像进行检测;而虹膜特征则通过霍夫梯度法进行检测,并利用区域增长法检测出虹膜,排除鼻孔误检情况,得到更加准确的特征。然后是图像融合模块,对于美瞳与输入图像融合,采用了加权平均法和IHS彩色空间转换法实现美瞳模板与输入图像的融合;对于框架眼镜与输入图像融合,则采用了直接复制法实现框架眼镜与输入图像的融合。最后通过Android开发技术,利用Activity等组件实现系统的界面和图像输入等,并且使用了 NDK开发技术,完成了对特征检测、图像融合本地C++代码的调用。实验结果表明,本文实现的面向智能终端的眼镜虚拟试戴系统取得了不错的试戴效果,能够满足用户的需求,解决了用户在购买眼镜过程中需要多次重复真实试戴等问题,有较好的实际应用价值。
[Abstract]:Nowadays, with the rapid development of computer and Internet technology, various software and business models emerge as the times require. At the same time, with the continuous improvement of hardware technology, the performance of intelligent terminals has changed by leaps and bounds. It plays an important role in people's daily life, and the number of users is increasing. With the rise of mobile Internet, "Internet" has become a new economic trend. The combination of traditional industries and the Internet has become a better way of development. More and more systems and software have been developed and applied in traditional industries. Eyeglasses are mainly divided into contact lenses and frame glasses, while contact lenses are loved by more and more young people because of their beautiful appearance. At present, the main problem is that it is not friendly and unsanitary for the user to choose the beauty pupil or the frame glasses, which need to be tried on many times. In order to solve this problem, this paper proposes to develop a software for the virtual glasses trial wear system for intelligent terminal. In order to improve the user's experience of buying glasses, the performance of intelligent terminals has been greatly improved, but the computing and processing ability can not be compared with that of computers in any case. So how to have low memory in smart terminals, The key to the development of this system is to realize the accurate location of human eye and iris features and the seamless fusion of different images under the condition of slow calculation. The work and main contents of this paper are divided according to the technical modules, which are the feature detection, respectively. Image fusion and Android development. Firstly, the human eye features are detected by AdaBoost algorithm, while the iris feature is detected by Hough gradient method, and iris is detected by region growth method. The image fusion module uses weighted averaging method and IHS color space conversion method to realize the fusion of the beauty pupil template and the input image. For the fusion of frame glasses and input image, direct copy method is adopted to realize the fusion of frame glasses and input image. Finally, through Android development technology, the interface and image input of the system are realized by using Activity and other components. Using NDK development technology, the feature detection and image fusion local C code are used. The experimental results show that the glasses virtual trial wear system oriented to intelligent terminal has achieved a good effect. It can meet the needs of the users and solve the problems that the users need to repeat the real wear in the process of purchasing glasses. It has good practical application value.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】