本文選題：視頻監(jiān)控 + 人數(shù)統(tǒng)計(jì)�。� 參考：《安徽大學(xué)》2014年碩士論文

【摘要】：目前學(xué)校教學(xué)樓內(nèi)照明等電子設(shè)備的開(kāi)關(guān)主要是依靠人工控制,從而造成經(jīng)常出現(xiàn)“無(wú)人亮燈”等情形。根據(jù)教室內(nèi)學(xué)生數(shù)量及分布,智能控制電力開(kāi)關(guān),可減少現(xiàn)實(shí)中電力資源大量浪費(fèi)的狀況,因而具有重要的意義。論文研究了基于視頻的人數(shù)統(tǒng)計(jì)問(wèn)題。隨著社會(huì)經(jīng)濟(jì)的發(fā)展,越來(lái)越多的視頻監(jiān)控系統(tǒng)廣泛應(yīng)用在各類公共場(chǎng)所。如何在這些視頻監(jiān)控?cái)?shù)據(jù)中提取出有用的信息,是目前該領(lǐng)域的研究熱點(diǎn)和難點(diǎn)之一。按照攝像頭與目標(biāo)物體的相對(duì)位置人數(shù)統(tǒng)計(jì)常見(jiàn)的場(chǎng)景有三種：攝像頭所處的位置位于目標(biāo)物體斜上方,能獲取目標(biāo)物體形狀輪廓信息、攝像頭位于進(jìn)出的通道口,能夠獲取目標(biāo)物體的頭部形狀輪廓信息以及攝像頭位于與目標(biāo)物體高度大概齊平的位置,能夠獲取臉部五官特征。這三種場(chǎng)景分別利用三種不同的特征實(shí)現(xiàn)統(tǒng)計(jì)功能,如利用人體的形狀特征、人的頭部輪廓特征以及臉部的五官特征。 由于教室內(nèi)的攝像機(jī)無(wú)法獲取完整的人體輪廓信息和清晰的人臉五官信息。本文做了如下工作： 1.本文將基于特征的方法用于教室場(chǎng)景中的人數(shù)檢測(cè),利用不同的特征提取方法與特征分類方法進(jìn)行不同的組合實(shí)現(xiàn)人數(shù)的統(tǒng)計(jì)。同時(shí),將稀疏表示的方法用于教室人數(shù)統(tǒng)計(jì)。并對(duì)這些方法進(jìn)行了實(shí)驗(yàn)驗(yàn)證以及比較分析。 2.本文針對(duì)教室內(nèi)學(xué)生大多幾乎處于靜止?fàn)顟B(tài),又偶有身體運(yùn)動(dòng)的特點(diǎn),提出了一種基于矩陣低秩稀疏分解的雙背景更新模型。先綜合利用幀間差分與低秩稀疏分解預(yù)判運(yùn)動(dòng)區(qū)域,并根據(jù)當(dāng)前視頻幀像素與預(yù)判的運(yùn)動(dòng)區(qū)域的位置關(guān)系,采用不同的參數(shù)更新背景模型；然后利用背景差分法獲取運(yùn)動(dòng)的目標(biāo)前景區(qū)域；最后對(duì)獲取的運(yùn)動(dòng)目標(biāo)區(qū)域進(jìn)行二值化、形態(tài)學(xué)處理、連通區(qū)域判斷等一系列操作,以消除噪聲影響并實(shí)現(xiàn)人數(shù)統(tǒng)計(jì)。 實(shí)驗(yàn)結(jié)果證明,本文提出的改進(jìn)的方法在教室人數(shù)統(tǒng)計(jì)中有一定的改善。最后,本文對(duì)研究過(guò)程中存在的不足進(jìn)行了分析,并對(duì)下一步工作計(jì)劃做出了闡述。
[Abstract]:At present, the switches of electronic devices such as lighting in school buildings rely on manual control, which often results in "unlit lights" and other situations. According to the number and distribution of students in the classroom, it is of great significance to control the power switch intelligently, which can reduce the waste of power resources in reality. This paper studies the number of people based on video. With the development of social economy, more and more video surveillance systems are widely used in all kinds of public places. How to extract useful information from these video surveillance data is one of the hot and difficult points in this field. According to the relative position of the camera and the target object, there are three common scenes: the camera is located in the position above the target object, it can obtain the contour information of the target object, and the camera is located at the entrance and exit of the target object. It can obtain the contour information of the head of the target object and the position of the camera which is about equal to the height of the target object, and can obtain facial features. The three scenes use three different features to achieve statistical functions, such as the shape of the human body, the contour of the human head and the facial features. Because the camera in the classroom can not obtain complete human contour information and clear facial features information. This paper has done the following work: 1. In this paper, the feature-based method is used to detect the number of students in the classroom scene, and different feature extraction methods and feature classification methods are used to realize the statistics of the number of students. At the same time, the sparse representation method is applied to the classroom population statistics. These methods are verified and compared with each other. 2. In view of the fact that most of the students in the classroom are in a static state and occasionally have the characteristics of physical motion, a dual background updating model based on matrix low rank sparse decomposition is proposed in this paper. First, the motion regions are predetermined by inter-frame difference and low-rank sparse decomposition, and different parameters are used to update the background model according to the position relationship between the current video frame pixels and the pre-determined moving regions. Then the background difference method is used to obtain the foreground region of the moving object. Finally, a series of operations such as binarization, morphological processing, connected region judgment and so on are carried out to eliminate the influence of noise and realize the statistics of the number of people. The experimental results show that the improved method proposed in this paper has a certain improvement in the statistics of classroom population. Finally, this paper analyzes the shortcomings of the research process, and describes the next work plan.
【學(xué)位授予單位】：安徽大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2014
【分類號(hào)】：TN948.6

【參考文獻(xiàn)】

相關(guān)期刊論文 前5條

1 郭景峰;劉漢峰;馬倩;;基于路網(wǎng)的移動(dòng)對(duì)象k近鄰查詢方法[J];計(jì)算機(jī)工程;2008年03期

2 周激流,張曄;人臉識(shí)別理論研究進(jìn)展[J];計(jì)算機(jī)輔助設(shè)計(jì)與圖形學(xué)學(xué)報(bào);1999年02期

3 錢鶴慶;陳剛;申瑞民;;基于人臉檢測(cè)的人數(shù)統(tǒng)計(jì)系統(tǒng)[J];計(jì)算機(jī)工程;2012年13期

4 張翠平,蘇光大;人臉識(shí)別技術(shù)綜述[J];中國(guó)圖象圖形學(xué)報(bào);2000年11期

5 白云飛;鄧建;董隴軍;李響;;深部硬巖巖爆預(yù)測(cè)的FDA模型及其應(yīng)用[J];中南大學(xué)學(xué)報(bào)(自然科學(xué)版);2009年05期

相關(guān)博士學(xué)位論文 前2條

1 張陽(yáng);復(fù)雜交通場(chǎng)景中基于視頻的行人檢測(cè)與跟蹤若干關(guān)鍵問(wèn)題研究[D];華南理工大學(xué);2013年

2 郭玲;智能視頻監(jiān)控中運(yùn)動(dòng)目標(biāo)檢測(cè)的算法研究[D];華南理工大學(xué);2013年

相關(guān)碩士學(xué)位論文 前1條

1 吳冰;基于人臉檢測(cè)的客流量統(tǒng)計(jì)研究[D];武漢科技大學(xué);2013年

，

本文編號(hào)：1824781