本文關(guān)鍵詞：數(shù)字視頻篡改檢測(cè)的被動(dòng)取證算法研究　出處：《浙江大學(xué)》2017年碩士論文　論文類型：學(xué)位論文

  更多相關(guān)文章： 篡改檢測(cè) 被動(dòng)取證 特征匹配 相似性度量 運(yùn)動(dòng)檢測(cè)

【摘要】：目前數(shù)字視頻在社會(huì)各領(lǐng)域應(yīng)用廣泛,視頻逐漸成為一種強(qiáng)大的信息傳輸媒介。隨著計(jì)算機(jī)技術(shù)的發(fā)展,各種視頻編輯軟件迅速普及,普通人無需專業(yè)知識(shí)即可對(duì)獲取視頻進(jìn)行有目的地篡改,在一些諸如法庭作證的應(yīng)用場(chǎng)合,篡改視頻可能會(huì)導(dǎo)致嚴(yán)重后果,破壞司法公正,影響社會(huì)穩(wěn)定。為了鑒定數(shù)字視頻的真實(shí)性與完整性,數(shù)字視頻取證技術(shù)逐漸成為多媒體信息安全領(lǐng)域中最為重要的研究課題之一,受到越來越多科研人員的關(guān)注。在數(shù)字視頻取證技術(shù)中,視頻篡改檢測(cè)研究開展最早,實(shí)際應(yīng)用價(jià)值也最為重大。當(dāng)前很多篡改檢測(cè)算法一般都需要提取原始幀與篡改幀各自特征進(jìn)行對(duì)比檢測(cè),即必須同時(shí)獲得原始視頻和篡改視頻,難以進(jìn)行有效的實(shí)際應(yīng)用。本文基于已有成熟的數(shù)字圖像取證方法和視頻獨(dú)有特征,針對(duì)視頻幀內(nèi)篡改和幀間篡改兩種主要的內(nèi)容篡改方式,聚焦于對(duì)待檢視頻直接盲檢測(cè)處理,根據(jù)提取的相關(guān)異常特征分析鑒別視頻真?zhèn)?以實(shí)現(xiàn)被動(dòng)取證研究。幀內(nèi)篡改在某一特定的視頻幀內(nèi)部操作,針對(duì)視頻幀內(nèi)復(fù)制-粘貼篡改操作,本文提出基于邊緣提取和特征點(diǎn)檢測(cè)匹配相結(jié)合的被動(dòng)取證方法。由于原始區(qū)域塊和復(fù)制區(qū)域塊之間具有很大的相似性,采樣視頻幀經(jīng)過預(yù)處理之后,本文采用邊緣檢測(cè)算子分析研究視頻幀中出現(xiàn)的相同邊緣線條。同時(shí),采用SIFT(Scale Invariant Features Transform,尺度不變特征變換)算法檢測(cè)提取采樣幀中的特征點(diǎn),基于余弦相似性度量方法,提出一種新的特征向量匹配方法以實(shí)現(xiàn)特征點(diǎn)的匹配聚類。相比較于2NN(2-Nearest-Neighbor,最近距離和次最近距離比值)方法,本文方法在速度性能上有明顯優(yōu)勢(shì)且檢測(cè)準(zhǔn)確率更高。實(shí)驗(yàn)結(jié)果表明本文被動(dòng)取證方法能有效檢測(cè)出視頻幀中存在的相同區(qū)域塊,辨別被復(fù)制區(qū)域的形狀和大小,并準(zhǔn)確地定位出克隆區(qū)域塊所在位置。粒度為幀的視頻幀插入、刪除和復(fù)制等幀間篡改操作會(huì)改變視頻幀的原始位置,本文提出基于時(shí)間域相關(guān)性分析的視頻幀間篡改被動(dòng)取證方法。采用HSV彩色直方圖作為視頻幀相似性度量特征,本文分別計(jì)算每一視頻幀的H-S二維直方圖和S-V二維直方圖,并進(jìn)行相鄰視頻幀的直方圖距離比較。根據(jù)直方圖距離出現(xiàn)的異常變化,本文方法能準(zhǔn)確地檢測(cè)出視頻幀插入、刪除和復(fù)制篡改�；诙ㄎ坏拇鄹奈恢�,利用特征相似性匹配,進(jìn)一步完成了篡改類型的取證復(fù)檢。視頻中存在的運(yùn)動(dòng)對(duì)象往往是人們重點(diǎn)關(guān)注的主體,本文提出基于運(yùn)動(dòng)學(xué)連續(xù)性分析的視頻幀間篡改被動(dòng)取證方法。視頻對(duì)象的運(yùn)動(dòng)學(xué)行為由真實(shí)運(yùn)動(dòng)所決定,但也會(huì)被篡改行為所改變。本文采用混合高斯模型背景建模和ViBe(Visual Background Extractor,視覺背景提取)算法對(duì)運(yùn)動(dòng)目標(biāo)進(jìn)行前景提取,通過提取運(yùn)動(dòng)目標(biāo)的異常運(yùn)動(dòng)軌跡,分析研究視頻幀刪除篡改;基于四鄰域搜索算法,本文計(jì)算運(yùn)動(dòng)目標(biāo)區(qū)域的質(zhì)心坐標(biāo),根據(jù)視頻目標(biāo)質(zhì)心參數(shù)的異常變化,成功檢測(cè)視頻幀復(fù)制篡改。同時(shí),根據(jù)運(yùn)動(dòng)軌跡和質(zhì)心參數(shù)異常變化出現(xiàn)位置,計(jì)算定位出相應(yīng)的篡改位置。
[Abstract]:At present, digital video widely used in various fields of the society, the video has gradually become a powerful information transmission media. With the development of computer technology, the rapid popularization of video editing software, ordinary people without professional knowledge to get the video to tampering, applications such as testifying in court, video tampering may cause serious consequences, destruction of justice, affect social stability. In order to identify the authenticity and integrity of digital video, digital video forensics technology has gradually become one of the most important research topics in the field of multimedia information security, has attracted more and more attention from researchers. In the digital video forensics technology, video tamper detection research on the earliest the actual application value is the most important. Many current tamper detection algorithms generally need to extract the original frame and the frame of tampering with each characteristic. Than that must be obtained from the original video detection, and tampering with the video at the same time, difficult to carry out effective practical application. This paper has a mature digital image forensics method and video based on the unique characteristics, for intra and inter frame video tampering tampering two main content tampering, focusing on the detected video direct blind detection processing, according to the characteristics of the relevant the anomaly extraction analysis to identify the authenticity of the video, in order to achieve passive forensics. Intraframe tampering in a particular video frame in the video frame for the internal operation, copy paste tampering, this passive forensics method of edge extraction and feature point detection and matching based on combination. Because of the great similarity between the original block copy and block sampling, video frame after pretreatment, this paper uses edge detection operator to analysis the same edge lines appear on the video frame at the same time, Using SIFT (Scale Invariant Features Transform, the scale invariant feature transform) algorithm to detect feature points extraction sampling frames, cosine similarity measure method is proposed based on a new feature vector matching method to achieve the matching feature point clustering. Compared with 2NN (2-Nearest-Neighbor, the nearest and next nearest distance ratio method, this paper) the method has obvious advantages and higher detection accuracy in speed performance. The experimental results show that the passive forensics method can effectively detect the same region exists in the video frame, to identify the replication region of shape and size, and accurately locate the region location. Clone size of video frame insertion, original the location of the deletion and replication of inter frame tampering will change the video frame, this paper take the method of video frames with passive time domain correlation analysis based on using H. SV color histogram as video frame similarity features, the thesis calculates each video frame H-S two-dimensional histogram and S-V histogram comparison, histogram distance and adjacent video frames. According to the abnormal changes of the distance histogram, this method can accurately detect the video frame insert, delete and copy location tampering tampering. Based on the position, using feature similarity matching, completing a further tampering with the type of evidence review. Moving objects in the video there is often the main focus of attention, this passive forensics method tampering with the video frame continuity analysis of kinematics based on kinematics behavior. The video object is determined by the real movement, but also will be behavior with change. This paper uses the Gauss mixture model background modeling and ViBe (Visual Background Extractor, visual background extraction algorithm of moving target) Standard for foreground extraction, the abnormal trajectory of moving object extraction, analysis and research of video frame deletion forgery; search algorithm based on four neighborhood, this paper calculates the centroid coordinates of target region, according to the abnormal change of the video target centroid parameters, the successful detection of video frame copy tampering. At the same time, according to the motion trajectory and centroid position parameter changes the calculation, to locate the appropriate tampering position.

【學(xué)位授予單位】：浙江大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類號(hào)】：TP391.41

【相似文獻(xiàn)】

相關(guān)期刊論文 前10條

1 駱偉祺;黃繼武;丘國(guó)平;;魯棒的區(qū)域復(fù)制圖像篡改檢測(cè)技術(shù)[J];計(jì)算機(jī)學(xué)報(bào);2007年11期

2 張?chǎng)?李學(xué)明;;改進(jìn)的基于顏色濾波陣列特性的篡改檢測(cè)[J];計(jì)算機(jī)工程與應(yīng)用;2009年06期

3 王鑫;魯志波;;數(shù)字圖像拷貝移動(dòng)篡改檢測(cè)方法的分析[J];計(jì)算機(jī)工程與設(shè)計(jì);2009年12期

4 周治平;張小祥;;基于質(zhì)量評(píng)價(jià)量和方差分析的圖像篡改檢測(cè)[J];計(jì)算機(jī)工程;2011年09期

5 馬巧梅;鄧啟森;;基于交叉劃分的灰度圖像篡改檢測(cè)與恢復(fù)算法[J];微電子學(xué)與計(jì)算機(jī);2014年09期

6 張震;蘇白娜;喻宙;;一種新的圖像拷貝篡改檢測(cè)方法[J];鄭州大學(xué)學(xué)報(bào)(工學(xué)版);2012年06期

7 黃添強(qiáng);陳智文;蘇立超;鄭之;袁秀娟;;利用內(nèi)容連續(xù)性的數(shù)字視頻篡改檢測(cè)[J];南京大學(xué)學(xué)報(bào)(自然科學(xué)版);2011年05期

8 黃添強(qiáng);曾文賦;;基于壓縮感知的視頻異源篡改檢測(cè)[J];計(jì)算機(jī)工程與應(yīng)用;2011年30期

9 周龍龍;;圖像空域分析下的篡改檢測(cè)算法[J];電視技術(shù);2012年23期

10 李曉梅;;基于殘留噪聲相關(guān)性的視頻篡改檢測(cè)算法[J];信息技術(shù);2013年10期

相關(guān)會(huì)議論文 前2條

1 孔輝;梁洪亮;辛陽;楊義先;陳林順;;一種高性能的網(wǎng)頁篡改檢測(cè)與恢復(fù)機(jī)制[A];2010年全國(guó)通信安全學(xué)術(shù)會(huì)議論文集[C];2010年

2 韓杰思;朱中梁;沈建京;彭韶峰;;針對(duì)拼接圖像篡改的檢測(cè)技術(shù)[A];第十四屆全國(guó)圖象圖形學(xué)學(xué)術(shù)會(huì)議論文集[C];2008年

相關(guān)博士學(xué)位論文 前3條

1 楊本娟;圖像篡改檢測(cè)中的投影核方法[D];貴州大學(xué);2015年

2 于立洋;數(shù)字圖像及視頻篡改檢測(cè)關(guān)鍵技術(shù)研究[D];哈爾濱工業(yè)大學(xué);2016年

3 丁琦;數(shù)字音頻篡改檢測(cè)與隱寫分析技術(shù)研究[D];解放軍信息工程大學(xué);2011年

相關(guān)碩士學(xué)位論文 前10條

1 張璐波;基于運(yùn)動(dòng)目標(biāo)的視頻幀內(nèi)篡改檢測(cè)算法研究與實(shí)現(xiàn)[D];上海交通大學(xué);2015年

2 劉躍強(qiáng);用于提高篡改檢測(cè)精度的半脆弱性數(shù)字水印技術(shù)研究[D];電子科技大學(xué);2014年

3 林志新;基于內(nèi)容的視頻篡改檢測(cè)系統(tǒng)設(shè)計(jì)與實(shí)現(xiàn)[D];電子科技大學(xué);2014年

4 孫艷茹;數(shù)字圖像區(qū)域復(fù)制粘貼篡改檢測(cè)算法分析與研究[D];黑龍江大學(xué);2015年

5 鄧峰;網(wǎng)頁防篡改中篡改檢測(cè)技術(shù)的研究[D];西安工業(yè)大學(xué);2013年

6 吳俞醒;基于連續(xù)性特征的視頻幀間篡改檢測(cè)算法的研究與實(shí)現(xiàn)[D];上海交通大學(xué);2015年

7 李君;基于客戶端的網(wǎng)頁篡改檢測(cè)研究[D];浙江工業(yè)大學(xué);2015年

8 王春井;基于網(wǎng)絡(luò)爬蟲與HASH的網(wǎng)站篡改檢測(cè)系統(tǒng)的設(shè)計(jì)與實(shí)現(xiàn)[D];內(nèi)蒙古大學(xué);2016年

9 劉海永;用于視頻篡改檢測(cè)的數(shù)字視頻取證技術(shù)[D];華南理工大學(xué);2016年

10 謝云芳;基于改進(jìn)SIFT的數(shù)字圖像篡改檢測(cè)及其FPGA實(shí)現(xiàn)[D];哈爾濱工業(yè)大學(xué);2016年

，

本文編號(hào)：1427333