發(fā)布時(shí)間：2021-08-13 21:43

　　信息的載體可以是語(yǔ)音信號(hào),圖像或者是視頻。相較于語(yǔ)音信號(hào),圖像或視頻中包含的信息要遠(yuǎn)遠(yuǎn)大于語(yǔ)音信號(hào)。如何從圖像或視頻中提取出有效信息就成為了一個(gè)熱門的研究話題。計(jì)算機(jī)視覺(jué)是機(jī)器學(xué)習(xí)的一個(gè)熱門方向,它的研究目標(biāo)是讓計(jì)算機(jī)能夠模擬人的大腦,對(duì)輸入的圖像和視頻進(jìn)行理解分析。計(jì)算機(jī)視覺(jué)的任務(wù)主要包括圖像分類,目標(biāo)識(shí)別,目標(biāo)跟蹤和語(yǔ)義分割。傳統(tǒng)的計(jì)算機(jī)視覺(jué)算法手工提取圖像特征,比如空間特征,顏色通道特征,頻率分布直方圖等。隨著深度學(xué)習(xí)研究的深入和并行計(jì)算硬件的提升,卷積神經(jīng)網(wǎng)絡(luò)開(kāi)始被廣泛應(yīng)用于計(jì)算機(jī)視覺(jué)算法中。圖像和視頻的一個(gè)主要的采集設(shè)備是攝像頭,而攝像頭主要的拍攝對(duì)象之一是行人。在實(shí)際應(yīng)用中,利用計(jì)算機(jī)視覺(jué)進(jìn)行行人的檢測(cè)與跟蹤,并在此基礎(chǔ)上進(jìn)行行人的行為統(tǒng)計(jì)和分析,可以給視頻監(jiān)控和人流管理規(guī)劃提供重要的技術(shù)支持和數(shù)據(jù)信息基礎(chǔ)。比如,在無(wú)人駕駛領(lǐng)域,行人檢測(cè)算法可以檢測(cè)出出現(xiàn)在車載攝像頭拍攝區(qū)域的行人;無(wú)人駕駛行車控制系統(tǒng)可以利用檢測(cè)得到的行人位置以及軌跡數(shù)據(jù)近一步進(jìn)行分析并調(diào)整行駛狀態(tài)和規(guī)劃路線。人流密集場(chǎng)景中,利用行人計(jì)數(shù)算法可以統(tǒng)計(jì)分析人流數(shù)量,并利用這些數(shù)據(jù)進(jìn)行安全預(yù)警和人流疏導(dǎo)。所以... 

【文章來(lái)源】：華中師范大學(xué)湖北省 211工程院校 教育部直屬院校

【文章頁(yè)數(shù)】：80 頁(yè)

【學(xué)位級(jí)別】：碩士

【文章目錄】：
Abstract
Chapter 1 Introduction
    1.1 Research Background and Significance
    1.2 Research Status
    1.3 Research Difficulties
    1.4 Main Content and Thesis Structure
    1.5 Conclusion
Chapter 2 Basic Principle
    2.1 Overview
    2.2 Principles Involved in Pedestrian Detection
        2.2.1 YOLO v3
        2.2.2 YOLO v3 Tiny
    2.3 Principles Involved in Pedestrian Tracking
        2.3.1 Intersection Over Union
        2.3.2 Hungarian Algorithm
        2.3.3 Kalman Filter
    2.4 Related Concepts of Pedestrian Tracking Based on MOT
    2.5 Conclusion
Chapter 3 Pedestrian Detection Algorithm Based on Mobilenet-Tiny
    3.1 Research Motivation
    3.2 Improved YOLO v3 Tiny Based on Depthwise Separable convolutional Filter
    3.3 Mobilenet-Tiny Model Training
        3.3.1 Build Data Set
        3.3.2 Mobilenet-Tiny Model Training
    3.4 Experiment and Analysis
    3.5 Conclusion
Chapter 4 Pedestrian Tracking Algorithm Based on Mobilenet-SORT
    4.1 Research Motivation
    4.2 Improved MOT Algorithm Mobilenet-SORT Based on SORT and Mobilenet-Tiny
        4.2.1 SORT
        4.2.2 Mobilenet-SORT
    4.3 Experimental Process and Result Analysis
        4.3.1 Data Processing
        4.3.2 Experiment Design
        4.3.3 Experimental Analysis
    4.4 Conclusion
Chapter 5 Pedestrian Counting Based on Mobilenet-SORT and Implementation on Rasp-berry Pi
    5.1 Research Motivation
    5.2 Pedestrian Counting Based on Mobilenet-SORT
        5.2.1 SORT Algorithm Counting Error Analysis Experiment
        5.2.2 Mobilenet-SORT Pedestrian Counting Based on Regional Information
    5.3 Implementation of Pedestrian Counting on Raspberry Pi
    5.4 Conclusion
Chapter 6 Summary and Outlook
    6.1 Summary
    6.2 Outlook
References
致謝
Appendix A 中文摘要



本文編號(hào)：3341189