Visual Object Tracking from Correlation Filter to Deep Learning

The book focuses on visual object tracking systems and approaches based on correlation filter and deep learning. Both foundations and implementations have been addressed. The algorithm, system design and performance evaluation have been explored for three kinds of tracking methods including correlation filter based methods, correlation filter with deep feature based methods, and deep learning based methods. Firstly, context aware and multi-scale strategy are presented in correlation filter based trackers; then, long-short term correlation filter, context aware correlation filter and auxiliary relocation in SiamFC framework are proposed for combining correlation filter and deep learning in visual object tracking; finally, improvements in deep learning based trackers including Siamese network, GAN and reinforcement learning are designed. The goal of this book is to bring, in a timely fashion, the latest advances and developments in visual object tracking, especially correlation filter and deep learning based methods, which is particularly suited for readers who are interested in the research and technology innovation in visual object tracking and related fields. Preface Acknowledgements Contents About the Authors 1 Introduction 1.1 Motivation and Challenge 1.2 Basic Concepts and Features 1.3 Evolution of Visual Object Tracking Technology 1.4 Chapter Outline References 2 Algorithms Foundations 2.1 Correlation Filter Basics 2.1.1 MOSSE 2.1.2 Discriminative Correlation Filter 2.1.3 Kernel Correlation Filter 2.2 Typical Deep Learning Model for Tracking 2.2.1 Convolutional Neural Networks Based Model 2.2.2 Siamese Networks Based Model 2.2.3 Generative Adversarial Networks Based Model 2.2.4 Reinforcement Learning Based Model 2.3 Performance Evaluation 2.3.1 Performance Evaluation Criteria 2.3.1.1 Evaluate the Accuracy of the Tracking Algorithm 2.3.1.2 Evaluate the Success Rate of the Tracking Algorithm 2.3.1.3 Other Evaluation Criteria for Tracking Algorithms 2.3.2 Benchmark Datasets 2.4 Summary References 3 Correlation Filter Based Visual Object Tracking 3.1 Introduction 3.2 Correlation Filter Tracker with Context Aware Strategy 3.2.1 Context Aware Strategy 3.2.2 Adaptive Update Model 3.2.3 Framework and Procedure 3.2.4 Experimental Results and Discussions 3.3 Correlation Filter Tracker with Scale Pyramid 3.3.1 Scale Pyramid Filter 3.3.2 Rich Image Feature Representation 3.3.3 Framework and Procedure 3.3.4 Experimental Results and Discussions 3.4 Correlation Filter Tracker with Multi-Scale Superpixels 3.4.1 Multi-Scale Superpixels Segmentation 3.4.2 Structure Based Optimization Strategy 3.4.3 Framework and Procedure 3.4.4 Experimental Results and Discussions 3.5 Summary References 4 Correlation Filter with Deep Feature for Visual Object Tracking 4.1 Introduction 4.2 Long-Short Term Correlation Filter Based Visual Object Tracking 4.2.1 Fusion of Deep Features and Hand-Crafted Features 4.2.2 Correlation Filters with Long-Short Term Update 4.2.3 Framework and Procedure 4.2.4 Experimental Results and Discussions 4.3 Context-Aware Correlation Filter Network 4.3.1 Context-Aware Correlation Filter Network 4.3.2 Channel Attention Mechanism 4.3.3 Update with High Confidence Strategy 4.3.4 Framework and Procedure 4.3.5 Experimental Results and Discussions 4.4 Auxiliary Relocation in SiamFC Framework 4.4.1 Auxiliary Relocation with Correlation Filters 4.4.2 Switch Function for SiamFC Framework 4.4.3 Framework and Procedure 4.4.4 Experimental Results and Discussions 4.5 Summary References 5 Deep Learning Based Visual Object Tracking 5.1 Introduction 5.2 Attention Shake Siamese Based Visual Object Tracking 5.2.1 Attention Mechanisms in Siamese Network 5.2.2 Shake-Shake Mechanism in Siamese Network 5.2.3 Framework and Procedure 5.2.4 Experimental Results and Discussions 5.3 Frequency-Aware Siamese Network Based Visual Object Tracking 5.3.1 Frequency-Aware Siamese Network 5.3.2 Pre-training and Joint Update 5.3.3 Framework and Procedure 5.3.4 Experimental Results and Discussions 5.4 Improved Generative Adversarial Network Based Visual Object Tracking 5.4.1 Improved Adversarial Learning Strategy 5.4.2 Precise ROI Pooling for Faster Feature Extraction 5.4.3 Framework and Procedure 5.4.4 Experimental Results and Discussions 5.5 Improved Policy-Based Reinforcement Learning Based Visual Object Tracking 5.5.1 Non-Convex Optimized Variance Reduced Backward Propagation 5.5.2 ε-Greedy Strategy for Action Space Exploration 5.5.3 Regression Based Reward Function 5.5.4 Framework and Procedure 5.5.5 Experimental Results and Discussions 5.6 Summary References 6 Summary and Future Work 6.1 Summary 6.2 Future Work