本文選題：小視頻檢索 + 分布式存儲��； 參考：《西安電子科技大學(xué)》2014年碩士論文

【摘要】：近幾年來，隨著“平安城市”建設(shè)的推廣，在公安刑偵等安防領(lǐng)域業(yè)務(wù)中視頻監(jiān)控系統(tǒng)得到了廣泛應(yīng)用。隨之而來的是監(jiān)控視頻信息的快速增長，如何對海量的監(jiān)控視頻數(shù)據(jù)進(jìn)行有效的存儲和檢索成為視頻監(jiān)控檢索研究領(lǐng)域里的熱點問題。 本文針對以上需求，在對監(jiān)控視頻數(shù)據(jù)和云計算技術(shù)研究的基礎(chǔ)上，設(shè)計并實現(xiàn)了一種基于HBase的海量小視頻存儲和檢索系統(tǒng)。考慮到用戶真正關(guān)注的是海量視頻中包含目標(biāo)人物執(zhí)行某些動作的視頻片段，為了更好的滿足用戶的需求，將系統(tǒng)的輸出設(shè)計為目標(biāo)人物執(zhí)行特定動作的視頻片段。 系統(tǒng)首先根據(jù)MPEG-7標(biāo)準(zhǔn)，利用MPEG-7描述定義語言和標(biāo)準(zhǔn)描述子對視頻數(shù)據(jù)的圖像特征進(jìn)行描述，并以XML文件格式輸出。在對XML進(jìn)行編碼解析后，，通過一個數(shù)據(jù)映射模型將XML數(shù)據(jù)的樹形結(jié)構(gòu)映射到HBase的面向列的數(shù)據(jù)模型中。在數(shù)據(jù)檢索模塊中，將事件發(fā)生的時間、地點、目標(biāo)人物的圖像、目標(biāo)人物的動作信息作為輸入，通過與視頻庫、視頻片段庫以及圖片信息庫中的值進(jìn)行匹配，最后按照相似度的大小順序輸出目標(biāo)視頻片段。本文設(shè)計了一組實驗，并通過實驗驗證了本文提出的海量小視頻存儲和檢索系統(tǒng)的性能和有效性。
[Abstract]:In recent years, with the promotion of the construction of "Ping an City", video surveillance system has been widely used in the field of public security and criminal investigation. With the rapid growth of surveillance video information, how to effectively store and retrieve the massive surveillance video data has become a hot issue in the field of video surveillance retrieval. In this paper, based on the research of surveillance video data and cloud computing technology, a large amount of small video storage and retrieval system based on HBase is designed and implemented. Considering that what users really pay attention to is the video segment which contains some actions performed by the target person. In order to better meet the needs of the user, the output of the system is designed as the video segment of the target person to perform certain actions. Firstly, according to MPEG-7 standard, the system uses MPEG-7 description definition language and standard descriptor to describe the image features of video data, and outputs them in XML file format. After encoding and parsing XML, the tree structure of XML data is mapped to column oriented data model of HBase by a data mapping model. In the data retrieval module, the image of the target person, the image of the target person and the action information of the target person are used as input in the data retrieval module, which matches the values in the video library, the video segment library and the picture information base. Finally, the target video segment is output according to the order of similarity. In this paper, we design a set of experiments, and verify the performance and effectiveness of the proposed system of mass small video storage and retrieval.
【學(xué)位授予單位】：西安電子科技大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2014
【分類號】：TN948.6;TP391.3

【參考文獻(xiàn)】

相關(guān)期刊論文 前2條

1 王振奇;;視頻數(shù)據(jù)存儲技術(shù)與發(fā)展趨勢[J];中國安防;2010年Z1期

2 鄭p

本文編號：2050865