声明
Abstract
摘要
Table of Contents
List of figures
List of Tables
Lists of Abbreviations
1 Introduction
1.1 Background
1.1.1 Terms definition
1.1.2 Areas covered
1.2 Motivation
1.3 Definition of the problem
1.4 Objectives and expected results
1.5 Research method
1.6 Dataset of Videos
1.7 Structure of the thesis
2 Theoretical Framework
2.1 Object Tracking Categorization
2.2 Corner Detector Combined with Optical Flow
2.2.1 Corner detection
2.2.2 Optical flow
2.3 Speeded up Robust Features (SURF)
2.4 Local Binary Pattern
2.5 Beyond Semi-Supervised Online Boosting Tracking
2.6 Mean Shift Tracking
2.7 Continuously Adaptive Mean Shift (CAMSHIFT) Tracking
2.7.1 Color Probability distribution and histogram back projection
2.7.2 Mass center calculation
2.7.3 CAMSHIFT advantages and disadvantages
2.8 CAMSHIFT/Mean-Shift Improvement in Literatures
2.8.1 CAMSHlFT and Mean-Shift combined with interest points
2.8.2 CAMSHIFT improvement using new HSV model
2.8.3 CAMSHIFT with improvement of object localization
2.9 Kalman filter
2.9.1 Mathematical Formulation of Kalman Filter
2.9.2 How does Kalman Filter work?
2.9.3 Example of appling the Kalman filter
2.9.4 Opencv and Kalam Filter
2.10 Summary
3 Adaptive CAMSHIFT for Multi-object Tracking using MDCM
3.1 The Key Steps of Multi-Dominant Color Model
3.2 Techniques of Multi-Dominant Color Model
3.2.1 Object Localization
3.2.2 Object Modeling
3.2.3 Making Color Mask
3.2.4 Segmentation
3.2.5 Histogram Back Projection
3.2.6 Result Analysis Method
3.3 Experiments Results
3.3.1 Implementations
3.3.2 Results and Discussions
3.3.3 Result analysis
3.4 Summary
3.4.1 Advantages
3.4.2 Some limitations
4 Adaptive Kalman Filter for Multi-Object Tracking Using MCMP
4.1 The keys steps of Multi-Color Model with Prediction
4.2 Techniques of multi-color model with Prediction
4.2.1 Image Acquisition
4.2.2 Color extraction
4.2.3 Thresholding
4.2.4 Median smoothing
4.2.5 Image center point detection and association
4.2.6 Kalman Filtering
4.2.7 Result analysis
4.3 Experiments Results
4.3.1 Real time detection
4.3.2 Real time Multi-Objects detection and tracking
4.3.3 Result analysis (test video from USB camera:60 Frames)
4.4 Summary
4.4.1 Advantages
4.4.2 Disadvantages
5 Conclusion and Further Work
5.1 Conclusion
5.2 Further work
References
Acknowledgements
中南大学;