无人机航拍视频中高精度车辆轨迹提取图像处理方法研究
发布时间：2022-01-01 15:27
车辆轨迹数据中包含丰富的交通运行和车辆行驶特性,对于交通流理论分析和建模起到了关键支撑。近年来,无人机航拍成为一种高效、便捷、经济的交通视频采集手段。本研究旨在构建从无人机航拍视频中进行高精确度车辆行驶轨迹提取的图像处理方法,主要包括（1）采用Canny边缘检测法对车辆进行检测识别;（2）基于相关内核滤波器（KCF）的车辆跟踪与数据校验;（3）提出了一种基于绝对变换差和（SATD）的车辆检测过程图像稳定方法。结果表明,本研究提出的方法可使检测正确率达到451（72.6%）,相比现有算法具有明细提升。此外,在获取的数据集中有368条（69.6%）轨迹存在较为显著的检测与跟踪误差。为了提高数据的可靠性以进一步完善数据库的建立,本研究对轨迹数据进行了部分手动跟踪,最终总共获取了621条车辆的完整轨迹数据。
【文章来源】：东南大学江苏省 211工程院校 985工程院校 教育部直属院校
【文章页数】：108 页
【学位级别】：硕士
【文章目录】：
摘要
abstract
ACKNOWLEDGEMENTS
Chapter1:INTRODUCTION
1.1.Background
1.2.Literature review
1.2.1.General trajectory extraction methods
1.2.2.Applications for image processing
1.2.3.Use of aerial videos in Transportation Engineering field and trajectory extraction
1.2.4.Studies based on vehicle trajectory databases
1.3.Purpose
1.4..Research contents
Chapter2:VIDEO RECORDING AND METHODOLOGICAL FRAMEWORK
2.1.Description of the location
2.2.Materials and tools applied
2.3.Video recording
2.4.Methodological framework
Chapter3:CANNY EDGE DETECTOR APPLIED TO VEHICLE DETECTION
3.1.Functioning of the Canny edge detector
3.2.Vehicle detection procedure
3.3.Limitations found during the vehicle detection process
3.3.1.Issues regarding the color of the vehicles
3.3.2.Issues regarding the light variation
3.3.3.Camera displacement and the practical solution
3.3.4.Problem on detecting lane-changers
3.4.Vehicle detection practical process
3.5.Noise occurrence in the vehicle detection process
3.5.1.Failed detection
3.5.2.Tilted trackers
3.5.3.Trackers when only half of the vehicle was detected
3.6.Results of the vehicle detection process
Chapter4:BLOCK MATCHING PROPOSED FOR THE CAMERA MOTION ISSUE
4.1.Functioning of Sum of Absolute Transformed Differences(SATD)method
4.2.A practical experiment results and limitations
4.2.1.Displacements measured from the Area
4.2.2.Displacements measured from the Area
4.3.Applying the method to the case study
4.4.Results and improvements compared to the existing vehicle detection method
Chapter5:KERNELIZED CORRELATION FILTER APPLIED TO VEHICLE TRACKING
5.1.Functioning of the object tracker algorithm
5.2.Difference between its proposed scope and the purpose of the present work
5.3.Limitations of the object tracker algorithm
5.3.0.Occlusion
5.3.1.Background clutter
5.3.2.Small displacements
5.3.3.Light variation
5.4.Vehicle tracking practical process
5.5.Types of noise in the vehicle tracking process
5.5.1.The tracker that loses the vehicle it was tracking
5.5.2.Instability of the vehicle tracker
5.6.Results of the vehicle tracking process
Chapter6:DATA CLEANSING,SMOOTHING AND VALIDATION
6.1.Visual checking and data cleansing
6.2.Raw data analysis and general information
6.3.Smoothing of the traveled distances
6.4.Smoothing results
6.5.Data validation
6.5.1.Macro analysis of the extracted data
6.5.2.Deeper analysis of the extracted data
Chapter7:DATABASE CONSTRUCTION
7.1.Database contents
7.2.Conventional variables
7.3.Data that was directly obtained from the video
7.4.Variables that demanded calculation to be obtained
7.4.1.Partial and accumulated distances
7.4.2.Velocity
7.4.3.Acceleration
7.5.The final database
Chapter8:CONCLUSION
8.1.Data extraction performance
8.2.Achievements
8.3.Future researches
8.4.Database publication and use
REFERENCES
Appendix A:DATABASE SUPPORT TABLE
Appendix B:DATABASE SAMPLE
本文编号：3562409
【文章来源】：东南大学江苏省 211工程院校 985工程院校 教育部直属院校
【文章页数】：108 页
【学位级别】：硕士
【文章目录】：
摘要
abstract
ACKNOWLEDGEMENTS
Chapter1:INTRODUCTION
1.1.Background
1.2.Literature review
1.2.1.General trajectory extraction methods
1.2.2.Applications for image processing
1.2.3.Use of aerial videos in Transportation Engineering field and trajectory extraction
1.2.4.Studies based on vehicle trajectory databases
1.3.Purpose
1.4..Research contents
Chapter2:VIDEO RECORDING AND METHODOLOGICAL FRAMEWORK
2.1.Description of the location
2.2.Materials and tools applied
2.3.Video recording
2.4.Methodological framework
Chapter3:CANNY EDGE DETECTOR APPLIED TO VEHICLE DETECTION
3.1.Functioning of the Canny edge detector
3.2.Vehicle detection procedure
3.3.Limitations found during the vehicle detection process
3.3.1.Issues regarding the color of the vehicles
3.3.2.Issues regarding the light variation
3.3.3.Camera displacement and the practical solution
3.3.4.Problem on detecting lane-changers
3.4.Vehicle detection practical process
3.5.Noise occurrence in the vehicle detection process
3.5.1.Failed detection
3.5.2.Tilted trackers
3.5.3.Trackers when only half of the vehicle was detected
3.6.Results of the vehicle detection process
Chapter4:BLOCK MATCHING PROPOSED FOR THE CAMERA MOTION ISSUE
4.1.Functioning of Sum of Absolute Transformed Differences(SATD)method
4.2.A practical experiment results and limitations
4.2.1.Displacements measured from the Area
4.2.2.Displacements measured from the Area
4.3.Applying the method to the case study
4.4.Results and improvements compared to the existing vehicle detection method
Chapter5:KERNELIZED CORRELATION FILTER APPLIED TO VEHICLE TRACKING
5.1.Functioning of the object tracker algorithm
5.2.Difference between its proposed scope and the purpose of the present work
5.3.Limitations of the object tracker algorithm
5.3.0.Occlusion
5.3.1.Background clutter
5.3.2.Small displacements
5.3.3.Light variation
5.4.Vehicle tracking practical process
5.5.Types of noise in the vehicle tracking process
5.5.1.The tracker that loses the vehicle it was tracking
5.5.2.Instability of the vehicle tracker
5.6.Results of the vehicle tracking process
Chapter6:DATA CLEANSING,SMOOTHING AND VALIDATION
6.1.Visual checking and data cleansing
6.2.Raw data analysis and general information
6.3.Smoothing of the traveled distances
6.4.Smoothing results
6.5.Data validation
6.5.1.Macro analysis of the extracted data
6.5.2.Deeper analysis of the extracted data
Chapter7:DATABASE CONSTRUCTION
7.1.Database contents
7.2.Conventional variables
7.3.Data that was directly obtained from the video
7.4.Variables that demanded calculation to be obtained
7.4.1.Partial and accumulated distances
7.4.2.Velocity
7.4.3.Acceleration
7.5.The final database
Chapter8:CONCLUSION
8.1.Data extraction performance
8.2.Achievements
8.3.Future researches
8.4.Database publication and use
REFERENCES
Appendix A:DATABASE SUPPORT TABLE
Appendix B:DATABASE SAMPLE
本文编号：3562409
本文链接：https://www.wllwen.com/kejilunwen/daoluqiaoliang/3562409.html
