三維空間中人體運(yùn)動速度估算方法的研究


頁數(shù)：28 字?jǐn)?shù)：10204


摘 要
隨著計(jì)算機(jī)視覺和圖像處理技術(shù)的不斷發(fā)展，基于視頻的人體運(yùn)動分析已成為計(jì)算機(jī)視覺領(lǐng)域一個備受關(guān)注的研究方向，其在智能監(jiān)控、人機(jī)交互、醫(yī)療診斷等方面都有著廣闊的應(yīng)用前景。運(yùn)動速度作為人體運(yùn)動信息的一個重要部分，是權(quán)衡一個人當(dāng)前運(yùn)動狀態(tài)的重要指標(biāo)。在特定的場合，人體運(yùn)動速度能反映出一個人的態(tài)度、情緒、潛在意圖、健康程度等信息。因此，對人體運(yùn)動速度的研究具有重要的現(xiàn)實(shí)意義。
本文對固定攝像機(jī)采集的視頻圖像序列進(jìn)行了基于運(yùn)動目標(biāo)檢測的人體運(yùn)動速度估算方法的研究。首先，采用三連幀差與背景差相結(jié)合的方法對視頻中的運(yùn)動人體進(jìn)行了提取；其次，提取運(yùn)動人體質(zhì)心特征進(jìn)行運(yùn)動目標(biāo)定位；最后，分析人體行走時(shí)質(zhì)心縱向運(yùn)動軌跡，獲取人體行走一步所花時(shí)間來估算人體運(yùn)動速度。
實(shí)驗(yàn)結(jié)果表明，本文所采用的三連幀差分與背景差分相結(jié)合進(jìn)行運(yùn)動目標(biāo)檢測的算法速度快，具有一定的魯棒性；根據(jù)人體質(zhì)心在垂直方向的運(yùn)動軌跡所具有的周期性來描述人體的運(yùn)動速度，通過波形分析法鑒別人體運(yùn)動周期的谷底，所得到的經(jīng)驗(yàn)閾值符合人體生理學(xué)的特征。
關(guān)鍵詞：速度估算；運(yùn)動檢測；質(zhì)心；特征提取

Abstract

With the development of computer vision and image processing technology, human motion analysis based on video has become a hot field, and it has broad application prospects in intelligent control, human-computer interaction, medical diagnosis,etc. As an important part of the motion information, velocity is a key factor for measuring the current state of motion. In special environment, velocity can reflect the information about person's attitude, mood, potential intentions, and the degree of health. Therefore, the research of human motion speed has important practical significance.
Based on motion detection in the video image sequence captured by fixed cameras, an approach to estimating human motion speed is studied in this thesis. Firstly, method that employs adaptive background subtraction and difference between three frames is used to extract human body in the video; Furthermore, centroid is extracted for locating target; Finally, based on analysis of centroid trajectory in the vertical direction, the time spent for walking a step is used to estimate human motion speed.
Experimental results show that the algorithm that employs adaptive background subtraction and difference between three frames is fast, and it has a certain degree of robustness; Periodicity of centroid trajectory in the vertical direction is used to describe the motion speed, and the correct trough is identificated through the analysis of body movement waveform, and threshold gained by the experiment can reflect the characteristics of human physiology.

Keywords: velocity estimation;motion detection;centre of gravity;feature extraction

目 錄
摘 要 I
ABSTRACT II
第一章 緒論 1
1.1 研究目的與意義 1
1.2 國內(nèi)外研究現(xiàn)狀 1
1.3 主要工作 2
1.4 論文結(jié)構(gòu) 2
第二章 運(yùn)動目標(biāo)檢測方法介紹 3
2.1 幀差法 3
2.2 背景差法 3
2.3 光流法 5
第三章 基于運(yùn)動目標(biāo)檢測的人體運(yùn)動速度估算 7
3.1 運(yùn)動目標(biāo)檢測 7
3.1.1 背景重構(gòu)與更新策略 8
3.1.2 運(yùn)動目標(biāo)提取 10
3.1.3 噪聲消除 10
3.2 運(yùn)動速度估算 12
3.2.1 質(zhì)心提取 13
3.2.2 速度估算 13
第四章 實(shí)驗(yàn)結(jié)果與分析 16
4.1 運(yùn)動目標(biāo)提取 16
4.2 人體運(yùn)動速度分析 17
4.3 實(shí)驗(yàn)總結(jié)分析 20
結(jié) 論 21
參考文獻(xiàn) 22
謝 辭 24