本文關(guān)鍵詞： 視頻摘要 運(yùn)動目標(biāo)檢測 背景建模 目標(biāo)跟蹤　出處：《西安電子科技大學(xué)》2014年碩士論文　論文類型：學(xué)位論文

【摘要】：隨著微電子技術(shù)和計(jì)算機(jī)信息技術(shù)的飛速發(fā)展,視頻等海量數(shù)據(jù)不斷積累使得用戶對視頻的查找和處理越來越困難,這使人們不得不發(fā)覺和完善對視頻數(shù)據(jù)的處理的相關(guān)技術(shù)。如今有關(guān)視頻處理的視頻摘要技術(shù)在各行各業(yè)發(fā)揮著越來越重要的作用,特別是監(jiān)控方面的應(yīng)用,大大減少了工作人員的工作量。視頻數(shù)據(jù)的結(jié)構(gòu)是非線性化的,無法按照一般的處理方法對其處理,視頻摘要是解決此問題眾多方法之一。本文主要研究了視頻處理中有關(guān)視頻摘要的技術(shù),視頻摘要的主要過程有讀入視頻、背景建模、前景提取、運(yùn)動目標(biāo)跟蹤、目標(biāo)的時序與空間規(guī)劃、濃縮視頻的生成。針對這幾個過程中存在的關(guān)鍵技術(shù)進(jìn)行詳細(xì)的介紹。一方面在目標(biāo)檢測中的背景模型創(chuàng)建提出了改進(jìn)算法,另一方面根據(jù)改進(jìn)算法設(shè)計(jì)了一個視頻摘要系統(tǒng)。首先論文研究了運(yùn)動目標(biāo)檢測算法,根據(jù)算法的基本原理將其分為了四類,簡單的背景建模、基于統(tǒng)計(jì)信息建模、非參數(shù)核密度估計(jì)建模、非背景建模。對每一類中選擇了經(jīng)典算法進(jìn)行了介紹和實(shí)驗(yàn)。在進(jìn)行幀差法提取前景的時候提出了一種改進(jìn)方法,該方法能夠有效的減弱粗影問題。其次論文研究了目標(biāo)跟蹤方法,運(yùn)動目標(biāo)跟蹤是整個設(shè)計(jì)優(yōu)化性技術(shù),如果運(yùn)用的合適得當(dāng),可以讓系統(tǒng)更加智能人性化,本章主要介紹了卡爾曼濾波,Mean Shift方法。給出了算法的基本原理。論文針對視頻摘要最關(guān)鍵的技術(shù),創(chuàng)建背景模型,在平均背景法提出的移動平均背景法的基礎(chǔ)上提出了一種基于傳統(tǒng)移動平均算法的改進(jìn)算法,并且對該算法進(jìn)行了實(shí)驗(yàn)分析。最后基于vs2010平臺,設(shè)計(jì)并開發(fā)了視頻摘要系統(tǒng),給出了系統(tǒng)的主要模塊和流程圖,并且給出了實(shí)際監(jiān)控視頻測試結(jié)果對其進(jìn)行了相應(yīng)分析。
[Abstract]:With the rapid development of microelectronics and computer information technology, video and other massive data accumulation makes it more and more difficult for users to find and process video. This has forced people to discover and improve the related technologies for video data processing. Nowadays, video summarization technology about video processing is playing an increasingly important role in various industries, especially in the application of surveillance. The structure of the video data is nonlinear and cannot be processed according to the normal processing method. Video summarization is one of the many methods to solve this problem. This paper mainly studies the technology of video summarization in video processing. The main processes of video summarization include reading video, background modeling, foreground extraction, moving target tracking, etc. The timing and spatial planning of the target, the generation of condensed video. The key technologies in these processes are introduced in detail. On the one hand, an improved algorithm is proposed to create the background model in target detection. On the other hand, a video summary system is designed according to the improved algorithm. Firstly, this paper studies the moving target detection algorithm, and divides it into four categories according to the basic principle of the algorithm, simple background modeling, modeling based on statistical information. Non-parametric kernel density estimation modeling, non-background modeling. The classical algorithms are introduced and experimented in each class. An improved method is proposed to extract the foreground of frame difference method. This method can effectively reduce the coarse shadow problem. Secondly, this paper studies the target tracking method, moving target tracking is the whole design optimization technology, if it is properly used, it can make the system more intelligent and humanized. This chapter mainly introduces the Kalman filter mean Shift method, and gives the basic principle of the algorithm. Based on the moving average background method proposed by the average background method, an improved algorithm based on the traditional moving average algorithm is proposed, and the algorithm is analyzed experimentally. Finally, based on the vs2010 platform, a video summarization system is designed and developed. The main modules and flow charts of the system are given, and the test results of the actual surveillance video are analyzed.
【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】：TP391.41;TN948.6

【參考文獻(xiàn)】

相關(guān)期刊論文 前1條

1 胡閩;劉純平;崔志明;王朝暉;張書奎;;聚類差分圖像核密度估計(jì)前景目標(biāo)檢測[J];中國圖象圖形學(xué)報;2009年10期

，

本文編號：1535402