一種基于多類支持向量機(jī)模型提高人臉識(shí)別精度的方法
發(fā)布時(shí)間:2024-02-02 11:07
面部識(shí)別是人工智能和圖像處理(模式識(shí)別)領(lǐng)域的主要研究方向之一,已經(jīng)廣泛應(yīng)用于身份認(rèn)證、視頻監(jiān)控和生物檢測(cè)。因?yàn)榉墙佑|式、自然、方便并且可靠,面部識(shí)別已成為生物識(shí)別系統(tǒng)的一個(gè)普遍選擇。面部識(shí)別準(zhǔn)確率仍然有待提高,一直被當(dāng)做一個(gè)重要的研究領(lǐng)域。因此,本文研究的主要目標(biāo)就是提高人臉識(shí)別的準(zhǔn)確性,F(xiàn)有的一些基于支持向量機(jī)(SVM)的方法,能夠提高人臉識(shí)別精度。參考這些方法,本文提出了一種基于SVM的新方法,可以更加有效識(shí)別人臉。本文方法可以顯著提高人臉識(shí)別精度,并可以用作合適的SVM多類模型。為了解決多于兩個(gè)類別(多類)的問(wèn)題,可以把幾個(gè)二進(jìn)制值分類器組合在一起,已有多個(gè)這種SVM與多分類器結(jié)合的模型得到了有效應(yīng)用。本研究采用了組合式多分類器,即分類式糾錯(cuò)輸出碼(ECOC)分類器。本文方法采用方向梯度直方圖(HOG)特征提取技術(shù)來(lái)提取圖像相關(guān)特征。通過(guò)借助一對(duì)一編碼設(shè)計(jì)方法,利用多類SVM對(duì)圖像進(jìn)行分類。此外,ECOC用于將多類分類問(wèn)題簡(jiǎn)化為一組二進(jìn)制分類問(wèn)題,來(lái)降低錯(cuò)誤率。為了驗(yàn)證本文提出的人臉識(shí)別方法的有效性,ORL、YALE、JAFFE和自建數(shù)據(jù)庫(kù)分別被用來(lái)進(jìn)行測(cè)試。實(shí)驗(yàn)結(jié)果顯示,本文...
【文章頁(yè)數(shù)】:60 頁(yè)
【學(xué)位級(jí)別】:碩士
【文章目錄】:
中文摘要
abstract
CHAPTER1 INTRODUCTION
1.1 Background
1.2 Applications of Facial Recognition
1.3 Future of Facial Recognition
1.4 Thesis Statement
1.5 Motivation behind the Research
1.6 Aims of the Research
1.7 Objectives of the Research
1.8 Practical Significance of the Research
1.9 Research Methodology Selection
1.10 Structure of this Thesis
CHAPTER2 LITERATURE REVIEW
CHAPTER3 METHODOLOGY
3.1 Feature Extraction
3.1.1 Histogram of Oriented Gradients(HOG)
3.2 Classification
3.2.1 One-Versus-One Method
3.2.2 Error Correcting Output Codes(ECOC)
CHAPTER4 EXPERIMENTS AND RESULTS ANALYSIS
4.1 Experiment and Result Analysis on ORL database
4.2 Experiment and Result Analysis on YALE Face database
4.3 Experiment and Result Analysis on JAFFE database
4.4 Experiment and Result Analysis on Own-created database
4.5 Discussion
CHAPTER5 CONCLUSION AND FUTURE WORKS
5.1 Conclusion
5.2 Future Works
REFERENCES
Acknowledgements
Published Academic Papers
本文編號(hào):3892559
【文章頁(yè)數(shù)】:60 頁(yè)
【學(xué)位級(jí)別】:碩士
【文章目錄】:
中文摘要
abstract
CHAPTER1 INTRODUCTION
1.1 Background
1.2 Applications of Facial Recognition
1.3 Future of Facial Recognition
1.4 Thesis Statement
1.5 Motivation behind the Research
1.6 Aims of the Research
1.7 Objectives of the Research
1.8 Practical Significance of the Research
1.9 Research Methodology Selection
1.10 Structure of this Thesis
CHAPTER2 LITERATURE REVIEW
CHAPTER3 METHODOLOGY
3.1 Feature Extraction
3.1.1 Histogram of Oriented Gradients(HOG)
3.2 Classification
3.2.1 One-Versus-One Method
3.2.2 Error Correcting Output Codes(ECOC)
CHAPTER4 EXPERIMENTS AND RESULTS ANALYSIS
4.1 Experiment and Result Analysis on ORL database
4.2 Experiment and Result Analysis on YALE Face database
4.3 Experiment and Result Analysis on JAFFE database
4.4 Experiment and Result Analysis on Own-created database
4.5 Discussion
CHAPTER5 CONCLUSION AND FUTURE WORKS
5.1 Conclusion
5.2 Future Works
REFERENCES
Acknowledgements
Published Academic Papers
本文編號(hào):3892559
本文鏈接:http://sikaile.net/kejilunwen/shengwushengchang/3892559.html
最近更新
教材專著