3D Imaging for Safety and Security (eBook)

Contents 6
Contributing Authors 8
Preface 14
Part I Biometrics 17
Chapter 1 3D ASSISTED FACE RECOGNITION: A SURVEY 18
1. INTRODUCTION 18
2. 3D SENSING FOR FACIAL BIOMETRICS 20
3. 3D FACE MODELS 25
4. FACE REGISTRATION 26
5. USE OF 3D IN RECOGNITION 33
6. SUMMARY AND CONCLUSIONS 34
ACKNOWLEDGMENTS 35
REFERENCES 35
Chapter 2 A SURVEY ON 3D MODELING OF HUMAN FACES FOR FACE RECOGNITION 40
1. INTRODUCTION 40
2. DATA ACQUISITION AND REPRESENTATION 42
3. ACTIVE MODELING TECHNIQUES 46
4. PASSIVE MODELING TECHNIQUES 51
5. RECONSTRUCTION QUALITY AND COMPARISONS 66
6. USE OF 3D FACE MODELS IN FACE RECOGNITION 68
7. FUTURE TRENDS 70
8. CONCLUSION 72
ACKNOWLEDGEMENT 74
REFERENCES 74
Chapter 3 AUTOMATIC 3D FACE REGISTRATION WITHOUT INITIALIZATION 84
1. INTRODUCTION 84
2. RELATED WORK 86
3. GAUSSIAN FIELDS FOR 3D FACE REGISTRATION 89
4. EXPERIMENTAL ANALYSIS 93
5. CONCLUSIONS 106
6. ACKNOWLEDGEMENTS 106
REFERENCES 106
Chapter 4 A GENETIC ALGORITHM BASED APPROACH FOR 3D FACE RECOGNITION 110
1. INTRODUCTION 111
2. FRAMEWORK OF THE SYSTEM 112
3. 3D FACIAL MODEL CREATION 113
4. 3D FACIAL MODEL LABELING 115
5. GOOD FEATURE SELECTION USING A GA-BASED APPROACH 120
6. FACE MODEL MATCHING 123
7. EXPERIMENTS AND ANALYSIS 127
8. CONCLUSIONS AND FUTURE WORK 131
ACKNOWLEDGEMENT 132
REFERENCES 132
Chapter 5 STORY OF CINDERELLA 134
1. INTRODUCTION 134
2. UBIQUITOUS ISOMETRIES 135
3. FLAT EMBEDDING AND CANONICAL FORMS 137
4. SPHERICAL CANONICAL FORMS 138
5. GENERALIZED MULTIDIMENSIONAL SCALING 139
6. COMPARISON OF PHOTOMETRIC PROPERTIES 142
7. CONCLUSIONS 144
REFERENCES 145
Chapter 6 HUMAN EAR DETECTION FROM 3D SIDE FACE RANGE IMAGES 148
1. INTRODUCTION 148
2. RELATED WORK, MOTIVATION AND CONTRIBUTIONS OF THE CHAPTER 150
3. TEMPLATE MATCHING BASED EAR DETECTION 152
4. SHAPE MODEL BASED EAR DETECTION 157
5. EXPERIMENTAL RESULTS 161
6. CONCLUSIONS 168
REFERENCES 169
Part II Safety and Security Applications 172
Chapter 7 SYNTHETIC APERTURE FOCUSING USING DENSE CAMERA ARRAYS 174
1. INTRODUCTION 174
2. REFOCUSING WITH HOMOLOGIES 177
3. REAL-TIME SYNTHETIC FOCUS 181
4. CONCLUSIONS 184
ACKNOWLEDGEMENTS 185
REFERENCES 185
APPENDIX A 186
Chapter 8 DYNAMIC PUSHBROOM STEREO VISION 188
1. INTRODUCTION 188
2. DYNAMIC PUSHBROOM STEREO GEOMETRY 191
3. GAMMA-RAY LINEAR PUSHBROOM STEREO 196
4. DYNAMIC STEREO MOSAICS FROM VIDEO 202
5. CONCLUSIONS AND DISCUSSIONS 211
ACKNOWLEDGEMENTS 213
REFERENCES 213
Chapter 9 3D MODELING OF INDOOR ENVIRONMENTS 216
1. INTRODUCTION 217
2. TECHNIQUES FOR MODELING 3D SCENES 218
3. OVERVIEW 220
4. HARDWARE PLATFORM AND SENSORS 222
5. BUILDING THE 2D MAP BY SCAN MATCHING 224
6. GENERATION OF GEOMETRY 227
7. GENERATION OF TEXTURES 227
8. ACQUISITION OF ADDITIONAL 3D GEOMETRY USING STEREO 230
9. RESULTS AND CONCLUSION 236
ACKNOWLEDGMENTS 237
REFERENCES 237
Chapter 10 3D SITE MODELLING AND VERIFICATION 240
1. INTRODUCTION 240
2. TASK COMPLEXITY 241
3. SYSTEM DESCRIPTION 241
4. DATA COLLECTION 243
5. REFERENCE MODEL CONSTRUCTION 251
6. VERIFICATION 252
7. EXPERIMENTS AND RESULTS 255
8. CONCLUSIONS 261
REFERENCES 262
Chapter 11 UNDER VEHICLE INSPECTION WITH 3D IMAGING 264
1. INTRODUCTION 265
2. THE “SENSOR BRICK” ARCHITECTURE FOR ROBOTIC UNDER VEHICLE INSPECTION 269
3. 3D IMAGING AND VISUALIZATION FOR UNDER VEHICLE INSPECTION 273
4. AUTOMATION FOR THREAT DETECTION 281
5. CONCLUSIONS AND FUTURE DIRECTIONS 290
ACKNOWLEDGEMENTS 291
REFERENCES 291
Colour Plate Section 294
Index 322

Chapter 2 A SURVEY ON 3D MODELING OF HUMAN FACES FOR FACE RECOGNITION (P. 25)

S. Huq, B. Abidi, S. G. Kong, and M. Abidi
Imaging, Robotics, and Intelligent Systems, The University of Tennessee, Knoxville,
TN 37996-2100, USA.
Abstract:

In its quest for more reliability and higher recognition rates the face recognition community has been focusing more and more on 3D based recognition. Depth information adds another dimension to facial features and provides ways to minimize the effects of pose and illumination variations for achieving greater recognition accuracy. This chapter reviews, therefore, the major techniques for 3D face modeling, the first step in any 3D assisted face recognition system.

The reviewed techniques are laser range scans, 3D from structured light projection, stereo vision, morphing, shape from motion, shape from space carving, and shape from shading. Concepts, accuracy, feasibility, and limitations of these techniques and their effectiveness for 3D face recognition are discussed.

Key words:

3D face reconstruction, face recognition, laser range scanner, structured light, stereo vision, morphing, shape from shading, shape from motion.

1. INTRODUCTION

Biometrics-based techniques such as fingerprint and iris matching often require physical contact or cooperation of the user. Face recognition offers a reliable means for personal identification without requiring much of the participant’s cooperation.

Despite the success in various applications, recognizing human faces in an uncontrolled environment has remained largely unsolved. The appearance of a face, an inherently three-dimensional (3D) object, projected onto a two-dimensional Face recognition has become one of the most active research fields in (2D) space is sensitive to the variations in pose and illumination.

Even, face variations of the same person created by pose and illumination changes could become larger than variations between individuals1. Face recognition techniques assisted with 3D facial models promise to overcome the difficulties and limitations associated with face recognition in 2D space.

In addition to being an important additional feature in recognition, depth plays a crucial role in mitigating variations caused by pose and illumination. In the acquisition of facial images using cameras deployed for surveillance or access control, people move freely with their faces appearing at any angle.

In this case, a smart face recognition system should be able to reproduce the same 2D rendering of the face as in the database for an accurate comparison. Once modeled in 3D, faces can accurately be back projected at any angle for matching.

The major goal of this chapter is to present a comprehensive review of the current technologies for modeling 3D human faces. Several related topics such as (1) the use of 3D face images in face recognition, (2) existing 3D face databases (publicly available), and (3) future trends of 3D face modeling and recognition have also been covered.

3D face modeling techniques can be divided into two classes - active and passive - depending on the imaging modalities and reconstruction methods.