Bøker av Zhengyou Zhang

1. Introduction.- 1.1 Brief Overview of Motion Analysis.- 1.2 Statement of the "Motion from Stereo" Problem.- 1.3 Organization of The Book.- 2. Uncertainty Manipulation and Parameter Estimation.- 2.1 Probability Theory and Geometric Probability.- 2.2 Parameter Estimation.- 2.2.1 Standard Kalman filter.- 2.2.2 Extended Kalman filter.- 2.2.3 Discussion.- 2.2.4 Iterated ExtendKalman Filter.- 2.2.5 Robustness and Confidence Procedure.- 2.3 Summary.- 2.4 Appendix: Least-Squares Techniques.- 3. Reconstruction of 3D Line Segments.- 3.1 Why 3D Line Segments.- 3.2 Stereo Calibration.- 3.2.1 Camera Calibration.- 3.2.2 Epipolar Constraint.- 3.3 Algorithm of the Trinocular Stereovision.- 3.4 Reconstruction of 3D Segments.- 3.5 Summary.- 4. Representations of Geometric Objects.- 4.1 Rigid Motion.- 4.1.1 Definition.- 4.1.2 Representations.- 4.2 3D Line Segments.- 4.2.1 Previous Representations and Deficiencies.- 4.2.2 A New Representation.- 4.3 Summary.- 4.4 Appendix: Visualizing Uncertainty.- 5. A Comparative Study of 3D Motion Estimation.- 5.1 Problem Statement.- 5.1.1 Line Segment Representations.- 5.1.2 3D Line Segment Transformation.- 5.2 Extended Kalman Filter Approaches.- 5.2.1 Linearization of the Equations.- 5.2.2 Derivation of Rotation Matrix.- 5.3 Minimization Techniques.- 5.4 Analytical Solution.- 5.4.1 Determining the Rotation.- 5.4.2 Determining the Translation.- 5.5 Kim and Aggarwal's method.- 5.5.1 Determining the Rotation.- 5.5.2 Determining the Translation.- 5.6 Experimental Results.- 5.6.1 Results with Synthetic Data.- 5.6.2 Results with Real Data.- 5.7 Summary.- 5.8 Appendix: Motion putation Using the New Line Segment Representation.- 6. Matching and Rigidity Constraints.- 6.1 Matching as a Search.- 6.2 Rigidity Constraint.- 6.3 Completeness of the Rigidity Constraints.- 6.4 Error Measurements inn the Constraints.- 6.4.1 Norm Constraint.- 6.4.2 Dot-Product Constraint.- 6.4.3 Triple-Product Constraint.- 6.5 Other Formalisms Rigidity Constraints.- 6.6 Summary.- 7. Hypothesize-and-Verify Method for Two 3D View Motion Analysis.- 7.1 General Presentation.- 7.1.1 Search in the Transformation Space.- 7.1.2 Hypothesize-and-Verify Method.- 7.2 Generating Hypotheses.- 7.2.1 Definition and Primary Algorithm.- 7.2.2 Control Strates in Hypothesis Generation.- 7.2.3 Additional Constraints.- 7.2.4 Algorithm of Hypothesis Generation.- 7.3 Verifying Hypothesis.- 7.3.1 Estimating the Initial Rigid Motion.- 7.3.2 Propagating Hyphoteses.- 7.3.3 Choosing the Best Hypothesis.- 7.3.4 Algorithm of Hypothesis Verification.- 7.4 Matching Noisy Segments.- 7.4.1 Version 1.- 7.4.2 Version 2.- 7.4.3 Version 3.- 7.5 Experimental Results.- 7.5.1 Indoor Scenes with a Large Common Part.- 7.5.2 Indoor Scenes with a Small Common Part.- 7.5.3 Rock Scenes.- 7.6 Summary.- 7.7 Appendix: Transforming a 3D Line Segment.- 8. Further Considerations on Reducing Complexity.- 8.1 Sorting Data Features.- 8.2 "Good-Enough" Method.- 8.3 Speeding Up the Hypothesis Generation Process Through Grouping.- 8.4 Finding Clusters Based on Proximity.- 8.5 Finding Planes.- 8.6 Experimental Results.- 8.6.1 Grouping Results.- 8.6.2 Motion Results.- 8.7 Conclusion.- 9. Multiple Object Motions.- 9.1 Multiple Object Motions.- 9.2 Influence of Egomotion on Observed Object Motion.- 9.3 Experimental Results.- 9.3.1 Real Scene with Synthetic Moving Objects.- 9.3.2 Real Scene with a Real Moving Object.- 9.4 Summary.- 10. Object Recognition and Localization.- 10.1 Model-Based Object Recognition.- 10.2 Adapting the Motion-Determination Algorithm.- 10.3 Experimental Result.- 10.4 Summary.- 11. Calibrating a Mobile Robot and Visual Navigation.- 11.1 The INRIA Mobile Robot.- 11.2 Calibration Problem.- 11.3 Navigation Problem.- 11.4 Experimental Results.- 11.5 Integrating Motion Information from Odometry.- 11.6 Summary.- 12. Fusing Multiple 3D Frames.- 12.1 System Description.- 12.2 Fusing Segments from Multiple Views.- 12.2.1 Fusing General Primitives.- 12.2.2 Fusing Line Segments.- 12.2.3 Ex...

Face detection, because of its vast array of applications, is one of the most active research areas in computer vision. In this book, we review various approaches to face detection developed in the past decade, with more emphasis on boosting-based learning algorithms. We then present a series of algorithms that are empowered by the statistical view of boosting and the concept of multiple instance learning. We start by describing a boosting learning framework that is capable to handle billions of training examples. It differs from traditional bootstrapping schemes in that no intermediate thresholds need to be set during training, yet the total number of negative examples used for feature selection remains constant and focused (on the poor performing ones). A multiple instance pruning scheme is then adopted to set the intermediate thresholds after boosting learning. This algorithm generates detectors that are both fast and accurate. We then present two multiple instance learning schemes for face detection, multiple instance learning boosting (MILBoost) and winner-take-all multiple category boosting (WTA-McBoost). MILBoost addresses the uncertainty in accurately pinpointing the location of the object being detected, while WTA-McBoost addresses the uncertainty in determining the most appropriate subcategory label for multiview object detection. Both schemes can resolve the ambiguity of the labeling process and reduce outliers during training, which leads to improved detector performances. In many applications, a detector trained with generic data sets may not perform optimally in a new environment. We propose detection adaption, which is a promising solution for this problem. We present an adaptation scheme based on the Taylor expansion of the boosting learning objective function, and we propose to store the second order statistics of the generic training data for future adaptation. We show that with a small amount of labeled data in the new environment, the detector's performance can be greatly improved. We also present two interesting applications where boosting learning was applied successfully. The first application is face verification for filtering and ranking image/video search results on celebrities. We present boosted multi-task learning (MTL), yet another boosting learning algorithm that extends MILBoost with a graphical model. Since the available number of training images for each celebrity may be limited, learning individual classifiers for each person may cause overfitting. MTL jointly learns classifiers for multiple people by sharing a few boosting classifiers in order to avoid overfitting. The second application addresses the need of speaker detection in conference rooms. The goal is to find who is speaking, given a microphone array and a panoramic video of the room. We show that by combining audio and visual features in a boosting framework, we can determine the speaker's position very accurately. Finally, we offer our thoughts on future directions for face detection. Table of Contents: A Brief Survey of the Face Detection Literature / Cascade-based Real-Time Face Detection / Multiple Instance Learning for Face Detection / Detector Adaptation / Other Applications / Conclusions and Future Work