【作者】 郑亮；

【导师】 王晖；

【作者基本信息】 国防科学技术大学 ， 控制科学与工程， 2010， 硕士

【摘要】 随着战争信息化程度的不断提高,网络媒体逐渐成为军事打击和舆论战的重要载体和骨干力量。网络媒体具有覆盖面广、信息传播速度快、信息总量无限、交互性好等特性,使得网络逐渐成为舆论战的主战场。在网络舆论传播和渗透所使用的媒体类型中,数字图像日益受到舆论攻击者的重视,正越来越频繁地使用在网络舆论攻击中,严重影响了社会的稳定和发展。针对这个问题,使用数字图像举证技术,通过网络上相似图像的检索,可以得到图像式网络舆论攻击中所使用图像相同或相似的图像信息,在此基础上进行对比与分析,所得到的图像鉴别信息更加直观、易于民众的理解、可信度更高,将这些鉴别信息用于舆论反制可以取得较好的效果。因此本文重点研究了面向图像式网络舆论攻击的举证反制技术。归纳起来,论文所做的创新性工作主要包括以下三个方面:首先,针对数字图像鉴别技术在图像式网络舆论攻击的反制中存在的不足,通过对图像式网络舆论攻击典型模式的分析,提出了面向图像式网络舆论攻击的举证反制方法。根据舆论攻击中所使用图像的特点,研究了数字图像举证技术的基本概念、工作原理和工作流程,并且对数字图像举证相关技术的国内外研究现状进行了综述。其次,在网络舆论的传播过程中,公共事件图像在图像式网络舆论攻击中使用的最为频繁。为了及时检索到相关图像信息,针对公共事件图像的特点,本文提出了一种基于对象组合特征的公共事件图像检索方法,该方法利用传统图像搜索引擎的检索结果,通过对公共事件图像中感兴趣对象的挖掘与识别,提取公共事件图像中的对象组合特征,使用SVM(Support Vector Machine,支持向量机)构建公共事件图像分类器,分类结果可以过滤掉检索结果中大量的无用信息,有效的检索到网络上与公共事件相关的图像,最后,数字图像举证技术的核心是相似图像检索方法,本文分别研究了基于SIFT(Scale Invariant Feature Transform,尺度不变特征变换)特征的相似图像检索方法和基于对象的相似图像检索方法,在此基础上提出了一种基于SIFT特征与对象特征相结合的相似图像检索方法SOFC (SIFT and Object Feature Combined content based image retrieval ),实验证明,与单纯使用基于SIFT特征或对象语义特征的相似图像检索方法相比,该方法可以进一步提高相似图像检索的精度和效率。本文为实现图像式网络舆论攻击的反制提供了良好的理论和技术基础。更多还原

【Abstract】 As information technology used in the war continues to be improved, online media has become an important backbone carrier of military strike and public opinion warfare, the internet media has many good characteristics such as a wide coverage, a fast information dissemination speed, an unlimited amount, good interaction and so on, making the network become the main battlefield of public opinion warfare.Among the media types using for penetration and diffusion in the public opinion warfare, digital image have become an increasing public attention to the attackers, and been increasingly used in the network public opinion attack. This will have a serious impact on social stability and development. Through the network, we can employ the image-based quote technique to get the same or similar images which are used in the network public opinion attack. After making a comparison and analysis between the source images and retrieval results, we could take an example to identify the reality of images. This identification information will achieve remarkable effect since it is easy for the public to understand, also more believable and intuitionistic. To address this problem, we focus on the study of quote counterattack techniques in the network public opinion attack based on the digital images, and this innovative work in this paper includes the following three aspects:First, due to the shortcomings of digital image forensics method for image-based network public opinion counterattack, we propose a quote counterattack approach by analyzing the typical mode for image-based network public opinion attack. According to the characteristics of the images used in network public opinion attack, we investigate the quote technique including the basic concept, working principle and its workflow. The related work about the quote technique on digital image is also reviewed.Secondly, in the propagation process of network public opinion, the most common way is using the images of public events. In order to retrieve the relevant image information in time, this paper proposes an image retrieval method based on object combination feature, and the method employs the result of the conventional image searching engine, extracts the combination feature of public events image by mining and recognition of objects, and uses SVM (Support Vector Machine) to construct a public events image classifier. The classification results can filter out a large number of useless information which is retrieved by text-based image searching engine, and find the related images of public events on the network effectively.Finally, the core of digital image quote technique is similar image retrieval method. This paper takes into account similar image retrieval method based on SIFT (Scale Invariant Feature Transform) feature and object-based image retrieval method. And then, this paper proposes to integrate SIFT and object feature together, it presents a combined image retrieval method SOFC (SIFT and Object Feature Combined content based image retrieval). Experimental results display that, compared with the similar image retrieval methods based on SIFT or object features, the SOFC remarkably improves the precise and effectiveness of similar image retrieval result.This study presents a good theoretical and technical basis for the counterattack technique against the image-based network public opinion attack.更多还原