人物图像处理技术研究及其在Web端实现

发布时间：2018-01-07 19:24

本文关键词：人物图像处理技术研究及其在Web端实现　出处：《华侨大学》2017年硕士论文　论文类型：学位论文

【摘要】：自20世纪70年代,图像处理自成一门学科以来,图像处理技术逐渐走进了人们的日常生活。近年来,随着物联网技术的长足发展,图像作为物联网中重要的信息来源,越来越体现出其重要的应用价值,并成为该应用领域一项不可或缺的技术。目前,图像处理技术的实现主要依赖于算法及硬件支持,在图像处理领域,大量的科研人员都在从事各种算法的研究,并通过强大的图形工作站,GPU等硬件来实现高清图像的处理。通常,图形工作站及相应的GPU的价值不菲,为科研人员的研究形成比较大的入门门槛。同时一些需要使用图像处理技术的普通人员又无法真正理解算法,也无法具备相应的硬件条件。因此,如果能将图像处理算法移植到服务器端,并通过Web的人机界面来向用户提供图像处理服务,就可以实现图像处理算法,甚至是硬件资源的平台共享,这符合目前计算模式服务化的大趋势。伴随着互联网技术的快速发展以及人们对当代网络的新要求,HTML5应运而生,它的出现,正好为解决上述问题提供了一条有效途径。同时,由于现阶段实现的基于Web的图像处理服务大部分都偏重娱乐性,忽略了在线提供图像处理服务的学术价值和研究意义,因此,本文在充分研究服务端架构,HTML5标准以及部分图像处理算法的基础上,将部分人物图像处理技术移植到了服务端进行实现,为浏览器客户端用户提供人脸检测与跟踪和行人检测两部分的图像处理服务。此外,随着大数据时代的来临,云计算平台和大数据处理的出现将会为本文今后的研究工作提供强有力的技术力量。本文所完成的研究工作具体有以下几个方面:1.本文首先详细介绍了本文的研究背景及研究意义,表明了以HTML5为基础在Web上提供图像处理服务的紧迫性,其次,本文具体阐述了图像处理服务化及HTML5在国内外的研究现状,最后,本文重点描述了图像处理及其在Web上实现服务化时所需要的技术理论基础。2.本文利用HTML5 Canvas技术实现了基于Web的人脸检测与跟踪。通过将HTML5 Canvas技术和经过优化的人脸特征分类器、人脸检测与跟踪算法相结合,本文在Web上实现了人脸检测与跟踪。实验结果表明,所提方法不仅能有效地运行在支持HTML5的浏览器上,而且还满足了人脸检测与跟踪过程中的实时性要求,保证了人脸检测的检测率,具有更优的检测效果。3.本文提出了一种基于HOG和Adaboost-BP模型的行人检测方法。通过图像预处理快速选定行人候选区域,在此基础上,利用HOG特征与Adaboost-BP模型实现对测试样本图像的检测识别。实验表明,本文方法与基于HOG-SVM算法的行人检测方法相比具有更高的检测率、更低的误检率和漏检率,在多个行人数据库上的实验结果表明本文方法具有较好的检测效果。为了能在Web上提供有关行人检测的图像处理服务,本文将利用HTML5 WebSocket协议在Web上实现行人检测。本课题为基于Web提供图像处理服务做了有益的探索性研究,希望本文所完成的工作能够给该领域的研究者们带来一些帮助和启发。
[Abstract]:Since 1970s, since the image processing as a discipline, image processing technology has gradually entered people's daily life. In recent years, with the rapid development of networking technology, networking image as an important source of information, more and more shows its important application value, and become an integral part of the application technology. At present, the realization of image processing technology mainly depends on the algorithm and hardware support in the field of image processing, a large number of researchers are engaged in the study of various algorithms, and through the powerful graphics workstation, GPU hardware to achieve the processing of high-definition images. Usually, the workstation and the corresponding GPU value of money for the study of scientific research personnel form a relatively large threshold. At the same time, some need to use image processing technology and the common people can not really understand the algorithm, also cannot have the corresponding hardware Conditions. Therefore, if the image processing algorithm is transplanted to the server, and through the Web interface to provide users with image processing services, you can achieve the image processing algorithm, and hardware resource sharing platform, which is consistent with the trend of the current computing services. With the rapid development of Internet technology and people the new requirements of the contemporary network HTML5 came into being, it appears, it provides an effective way to solve the above problems. At the same time, because of the image processing Web services mostly focus on entertainment based on at present, ignoring the online academic value and significance of the research, image processing service so the server research on the framework of full, HTML5 standard and basic part of image processing algorithms, image processing technology will be part of the characters were transplanted to the server, to browse Is the client provides image processing service detection face detection and tracking of pedestrians and two parts. In addition, with the advent of the era of big data, cloud computing platform and large data processing will provide strong technical force for the future research work. The research work of this paper has done the following aspects: 1. in this paper first introduces the research background and the significance of this study, that is based on HTML5 to provide the urgency of image processing services in Web. Secondly, the paper describes the image processing service and HTML5 research status at home and abroad, finally, this paper describes the image processing and achieve the required service in Web on the theoretical basis of.2. technology using HTML5 Canvas technology to achieve the Web face detection and tracking based on HTML5 and Canvas technology. The optimized human face The feature classifier combining face detection and tracking algorithm, this paper realizes the face detection and tracking in Web. The experimental results show that the proposed method can not only effectively run in support of HTML5 browser, but also meet the real-time face detection and tracking in the process to ensure the detection rate of face detection and has better effect of.3. this paper proposes a pedestrian detection method based on HOG and Adaboost-BP model. Through image preprocessing fast pedestrian candidate regions selected, on the basis of this, the use of HOG features and Adaboost-BP model to achieve the test sample image detection and recognition. Experiments show that this method compared with pedestrian detection method HOG-SVM based on the algorithm has a higher detection rate, lower false detection rate and false negative rate, in a number of pedestrian database. The experimental results show that this method has better detection effect. Fruit. In order to provide image processing services on pedestrian detection in Web, this paper will use the HTML5 WebSocket protocol implementation of pedestrian detection in Web. This paper provides Web with image processing services to do a useful exploratory study based on the hope that the work of this thesis can give researchers in this field and bring some help inspired.

【学位授予单位】：华侨大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP393.09

【相似文献】