RESEARCH ON OBJECT DETECTION AND TRACKING ALGORITHMS IN INTELLIGENT VIDEO SURVEILLANCE
RESEARCH ON OBJECT DETECTION AND TRACKING ALGORITHMS IN INTELLIGENT VIDEO SURVEILLANCE
Nan Hu
Shenyang Jianzhu University
Yu Ying
Shenyang Jianzhu University
Tianbo Xu
Shenyang Jianzhu University
Copyright © 2021 by Cayley Nielson Press, Inc.
ISBN: 978-1-7348822-7-8
Cayley Nielson Press Scholarly Monograph Series Book Code No.: 211-10-3
US$175.60
Preface
The intelligent video surveillance system that aims to realize the preset monitoring tasks by using computer vision technology to analyze and understand images and video signals is one of the important technical projects in our country. In the intelligent video surveillance system, object detection and tracking are a basic and important surveillance task, which is also the research focus in the intelligent surveillance system and the research hotspot of researchers in related fields. Based on the analysis of domestic and overseas researches, a series of object detection and tracking problems in the intelligent video surveillance system are studied in depth in this dissertation, including foreground motion detection problems, pedestrian detection problems in thermal surveillance, and anomaly detection problems in surveillance scene.
As the deep neural networks make success in feature representation, the features extracted by a deep neural network represent the appearance and motion patterns in different scenes more specifically. This book solves the various problems in the object detection and tracking area for intelligent video surveillance system and promotes the development of computer vision. It has broad application prospect and brings significant economic and social benefits.
In chapter 1, an efficient anomaly detection system for crowded scenes using variational autoencoders is introduced. In chapter 2 is about the essential ones of implicit active Contours driven by local and global image fitting energy for image segmentation and targent localization. In chapter 3 is about the essential ones of the fully convolutional network and visual saliency-based automatic optic disc detection in retinal fundus images. In chapter 4, the essential one of studying is the object tracking algorithm based on bilateral structure tensor under lie group manifold. The chapter 5 is about the covariance tracking algorithm on bilateral filtering under lie group structure. The chapter 6 is about the corner detection based on bilateral structure under lie group. The chapter 7 is about the object tracking based on bilateral structure tensor. The last chapter is about the visual object tracking using particle filtering with dual manifold models.
I wanted to thank the authors of the book chapters, especially Hongwei Li, Rui Men, Wei Gong, Xiaoxi Tian, Rui Zhang, Meiju Liu, Jialin Sun, because they offered to the technology of the computer vision applied to many areas of the latest progress and the existing challenges. This work was supported by the Liaoning Province Science and Technology Research Projects lnqn202014.
Nan Hu
Shenyang Jianzhu University
Shenyang,Liaoning, China
August 30,2021
Contents
PREFACE 1
CHAPTER 1 AN EFFICIENT ANOMALY DETECTION SYSTEM FOR CROWDED SCENES USING VARIATIONAL AUTOENCODERS
1.1 INTRODUCTION 8
1.2 RELATED WORK 12
1.2.1 Hand-Crafted Features Based Method 12
1.2.2 Deep Learning Based Method 13
1.3 PROPOSED METHOD 15
1.3.1 Overall Scheme 15
1.3.2 Convolutional VAE Architecture and Feature Extraction 17
1.3 ANOMALY DETECTION AND LOCALIZATION 23
1.4 EXPERIMENTAL RESULTS AND COMPARISONS 25
1.4.1 Visualization of the Feature Distribution 26
1.4.2 Qualitative and Quantitative Results 28
1.4.3 Run-Time Analysis 35
1.5 CONCLUSION 36
CHAPTER 2 IMPLICIT ACTIVE CONTOURS DRIVEN BY LOCAL AND GLOBAL IMAGE FITTING ENERGY FOR IMAGE SEGMENTATION AND TARGENT LOCALIZATION
2.1 INTRODUCTION 38
2.2 CV AND LIF 41
2.2.1 C-V Model 41
2.2.2 LIF Model 43
2.3 THE PROPOSED METHOD 45
2.3.1 The Formulation of the Proposed Model 45
2.3.2 The Implementation of the Proposed Model 47
2.4 EXPERIMENTAL RESULTS 49
2.4.1 Comparisons with the C-V Model 50
2.4.2 Comparisons with the LIF Model 51
2.5 CONCLUSION 56
CHAPTER 3 FULLY CONVOLUTIONAL NETWORK AND VISUAL SALIENCY-BASED AUTOMATIC OPTIC DISC DETECTION IN RETINAL FUNDUS IMAGES
3.1 INTRODUCTION 58
3.2 BACKGROUND 66
3.2.1 Fully Convolutional Network 66
3.2.2 Saliency Detection via Single-layer Cellular Automata 67
3.3 THE PROPOSED METHOD 70
3.3.1 Optic Disc Region Localization 71
3.3.2 Visual Saliency-based OD Detection 73
3.3.3 Single layer cellular automata 75
3.4 EXPERIMENTAL RESULTS AND ANALYSIS 79
3.5 CONCLUSIONS 86
CHAPTER 4 OBJECT TRACKING ALGORITHM BASED ON BILATERAL STRUCTURE TENSOR UNDER LIE GROUP MANIFOLD
4.1 INTRODUCTION 87
4.2. BILATERAL FILTERING THEORY 89
4.3 ESTABLISH OF BILATERAL STRUCTURAL TENSOR MATRIX 93
4.4. PARTICLE FILTER ALGORITHM 94
4.5. TRACKING ALGORITHM DESIGN AND TEMPLATE UPDATE STRATEGY 97
4.5.1 Tracking algorithm Design 97
4.5.2 Template Update Strategy 98
4.6. EXPERIMENTAL RESULTS AND ANALYSIS 99
4.7 CONCLUSIONS 102
CHAPTER 5 COVARIANCE TRACKING ALGORITHM ON BILATERAL FILTERING UNDER LIE GROUP STRUCTURE
5.1 INTRODUCTION 104
5.2 BILATERAL FILTERING THEORY 106
5.3 COVARIANCE MATRIX MODELING 108
5.4 DISTANCE UNDER LOG-EUCLIDEAN METRIC 109
5.5 TRACKING METHOD AND UPDATE STRATEGY 111
5.5.1 Tracking method 111
5.5.2 Updating strategy 112
5.6. EXPERIMENTS AND RESULTS 113
5.7 CONCLUSIONS 117
CHAPTER 6 CORNER DETECTION BASED ON BILATERAL STRUCTURE UNDER LIE GROUP
6.1 INTRODUCTION 119
6.2 CONSTRUCTING BILATERAL STRUCTURE TENSOR 122
6.2.1 Harris algorithm 122
6.2.2 Building pixel structure information by bilateral structure tensor 124
6.3 BUILDING BILATERAL STRUCTURE TENSOR ACCORDING TO LOG-EUCLID MEAN 126
6.4 DETERMINGING THE FEATURE POINTS 127
6.5 EXPERIMENTAL RESULTS 130
6.6 CONCLUSIONS 133
CHAPTER 7 OBJECT TRACKING BASED ON BILATERAL STRUCTURE TENSOR
7.1 INTRODUCTION 135
7.2 BILATERAL STRUCTURE TENSOR 138
7.3 FAST CALCULATION OF BILATERAL STRUCTURE TENSOR 140
7.4 DISTANCE BETWEEN STRUCTURE TENSORS 141
7.5 TRACKING ALGORITHM AND MODEL UPDATING STRATEGY 143
7.5.1 Tracking algorithm 143
7.5.2 Model updating strategy 144
7.6 EXPERIMENTAL RESULTS 146
7.7 CONCLUSIONS 149
CHAPTER 8 VISUAL OBJECT TRACKING USING PARTICLE FILTERING WITH DUAL MANIFOLD MODELS
8.1 INTRODUCTION 150
8.2 TWO MANIFOLD MODELS 153
8.2.1 SL (3)group 153
8.2.2 Region Covariance Manifold 155
8.3 TRACKING MODEL 158
8.3.1 Particle Filter 158
8.3.2 Dynamic model 160
8.3.3 Observation model 161
8.4 TRACKING ALGORITHM 162
8.5 EXPERIMENTAL RESULTS 164
8.5.1 Experiments for geometric deformation 164
8.5.2 Experiments for obscured and illumination change 170
8.6 CONCLUSION 171
REFERENCES 173
Readership
This book should be useful for students, scientists, engineers and professionals working in the areas of optoelectronic packaging, photonic devices, semiconductor technology, materials science, polymer science, electrical and electronics engineering. This book could be used for one semester course on adhesives for photonics packaging designed for both undergraduate and graduate engineering students.
