A Few Steps Towards 3D Active Vision

ISBN-13: 9783642645808 / Angielski / Miękka / 2013 / 243 str.

Thierry Vieville

Opis

T. Vieville: A Few Steps Towards 3D Active Vision appears as Vol. 33 in the Springer Series in Information Sciences. A specific problem in the field of active vision is analyzed, namely how suitable is it to explicitly use 3D visual cues in a reactive visual task? The author has collected a set of studies on this subject and has used these experimental and theoretical developments to propose a synthetic view on the problem, completed by some specific experiments. With this book scientists and graduate students will have a complete set of methods, algorithms, and experiments to introduce 3D visual cues in active visual perception mechanisms, e.g. autocalibration of visual sensors on robotic heads and mobile robots. Analogies with biological visual systems provide an easy introduction to this subject."

Spis treści

1. From 2D to 3D Active Vision.- 1.1 The Concept of Active Vision.- 1.1.1 Can Reactive Vision Be “Better” Than Passive Vision?.- 1.1.2 A Blink at the State of the Art in Active Vision.- 1.2 A Short Review of Existing Active Visual Systems.- 1.2.1 Active Visual Sensors.- 1.2.2 Control for Active Vision.- 1.2.3 Auto-Calibration of an Active Visual System.- 1.2.4 Perception of 3D Parameters Using Active Vision.- 1.3 Architecture of an Active Visual System.- 1.3.1 The Three Main Functions of an Active Visual System.- 1.3.2 Basic Requirements of a Visual System.- 1.4 2D Versus 3D Vision in an Active Visual System.- 1.4.1 Active Vision and 2D Visual Servoing.- 1.4.2 Introduction of 3D Vision in Visual Loops.- 1.4.3 Basic Modules for a 3D Active Visual System.- 1.5 Gaze Control in 3D Active Visual Systems.- 2. 3D Active Vision on a Robotic Head.- 2.1 A One-to-One 3D Gaze Controller.- 2.1.1 Technical Data on the Robotic Head.- 2.1.2 Computing the Inverse Kinematic for 3D Fixation.- 2.2 Active Observation of a 3D Visual Target.- 2.2.1 A ID Model of the Eye-Neck Tracking.- 2.2.2 A Linearized Adaptive ID Model of the Eye-Neck Tracking.- 2.2.3 Statistical Filtering of the Linearized Model.- 2.2.4 Controlling the Neck and Eye Positions.- 2.2.5 Automatic Tuning from the System Residual Error.- 2.2.6 Estimating VR.- 2.2.7 Estimating ?.- 2.2.8 Estimating ?.- 2.2.9 Simulation of the Combined Behavior.- 2.3 Detection of Visual Targets for 3D Visual Perception.- 2.4 Computing the 3D Parameters of a Target.- 2.4.1 A Unique Framework to Integrate Depth from Vergence, Motion and Zoom.- 2.4.2 Using Second-Order Focus Variations to Compute Depth.- 2.4.3 Multi-Model Concurrency in 3D Tracking.- 2.4.4 Considering Further 3D Information.- 2.5 Experimental Results.- 2.5.1 Head Intrinsic Calibration.- 2.5.2 Looking at a 3D Point.- 2.5.3 Where to Look Next Experiment.- 2.5.4 Reconstruction of a Coarse 3D Map.- 2.5.5 Tracking of a 3D Target: Simulation Experiments.- 2.5.6 Tracking of a 3D Target: Real Object Experiments.- 2.5.7 Conclusion.- 3. Auto-Calibration of a Robotic Head.- 3.1 Introduction.- 3.2 Reviewing the Problem of Visual Sensor Calibration.- 3.3 Equations for the Tracking of a Stationary Point.- 3.4 Recovering the Parameters of the Trajectory.- 3.4.1 An Initial Estimate of the Coefficients.- 3.4.2 Minimizing the Nonlinear Criterion.- 3.5 Computing Calibration Parameters.- 3.5.1 Equations for the Intrinsic Calibration Parameters.- 3.5.2 Extrinsic Parameters Computation.- 3.5.3 Calibration Algorithm.- 3.6 Experimental Results.- 3.6.1 How Stable are Calibration Parameters When Zooming?.- 3.6.2 Experiment 1: Parameter Estimation with Synthetic Data.- 3.6.3 Experiment 2: Trajectory Identification Using Real Data.- 3.6.4 Experiment 3: Parameters Estimation Using Real Data.- 3.7 Conclusion.- 3.8 Comparison with the Case of Known Translations.- 3.9 Application to the Case of a Binocular Head.- 3.10 Instantaneous Equations for Calibration.- 3.10.1 Reviewing the Definition of the Fundamental Matrix.- 3.10.2 Characterizing the Essential Matrix for Fixed Axis Rotations.- 3.10.3 Calibration Using the Fundamental Matrix.- 3.10.4 Discussion.- 4. Inertial Cues in an Active Visual System.- 4.1 Introduction.- 4.2 The Use of Inertial Forces in a Robotic System.- 4.2.1 Origins of Inertial Forces on a Robot.- 4.2.2 Distinction with Inertial Navigation on Vehicles.- 4.2.3 Available Inertial Sensors.- 4.2.4 Comparison with the Human Vestibular System.- 4.3 Auto-Calibration of Inertial Sensors.- 4.3.1 Presentation.- 4.3.2 Sensor Models.- 4.3.3 Accelerometers Intrinsic Calibration.- 4.3.4 Static Evaluation of Calibration Parameters.- 4.3.5 Experimental Procedure for Accelerometers Calibration.- 4.3.6 Gyrometers Intrinsic Calibration.- 4.3.7 Principle of the Dynamic Calibration.- 4.3.8 Computing the Rotation from g.- 4.3.9 Performing a Rotation in the Vertical Plane.- 4.3.10 Experimental Procedure for Calibration.- 4.3.11 Experimental Results for Calibration.- 4.4 Separation Between Gravity and Linear Acceleration.- 4.4.1 Using Special Assumptions on the Environment.- 4.4.2 Method 1: Estimating ?(O) in an Absolute Frame of Reference.- 4.4.3 Method 2: Estimating ?(O) Using the Jerk.- 4.5 Integration of Angular Position.- 4.5.1 A Rational Representation of 3D Rotations.- 4.5.2 Relation Between ? and the Angular Position.- 4.5.3 Cooperation of the Inertial and Visual Systems.- 4.5.4 Application of Inertial Information in a Visual System.- 4.5.5 Computation of the Scale Factor.- 4.6 Computing Self-Motion with a Vertical Estimate.- 4.6.1 A Representation of Self-Motion Using Vertical Cues.- 4.6.2 The “Gyro-Rotation” and the Vertical Rectification.- 4.6.3 Stabilization Using 2D Translation and Rotation.- 4.6.4 Implications on the Structure from Motion Paradigm.- 4.6.5 Estimation of the 3D Rotation.- 4.6.6 Implication on Token Matching.- 4.6.7 Experimental Results.- 4.7 Conclusion.- 5. Retinal Motion as a Cue for Active Vision.- 5.1 Definition and Notation.- 5.1.1 Calibration: The Camera Model.- 5.1.2 Motion: Discrete Rigid Displacements.- 5.1.3 Structure: Rigid Planar Structures.- 5.2 Using Collineations to Analyse the Retinal Motion.- 5.2.1 Definition.- 5.2.2 Properties.- 5.3 Estimation of Retinal Motion from Correspondences.- 5.3.1 Estimation of a H-Matrix from Point Correspondences.- 5.3.2 Performing Tests on the Estimate.- 5.3.3 Considering Correspondences Between Non-punctual Primitives.- 5.4 Implementation and Experimental Results.- 5.5 Conclusion.- 6. Uncalibrated Motion of Points and Lines.- 6.1 Introduction.- 6.2 Representations of the Retinal Motion for Points.- 6.2.1 Considering the Euclidean Parameters of the Scene.- 6.2.2 From Euclidean to Affine Parameters.- 6.2.3 Prom Affine to a Projective Parameterization.- 6.2.4 Conclusion: Choosing a Composite Representation.- 6.3 Representations of the Retinal Motion for Points and Lines.- 6.3.1 The Retinal Motion for Lines.- 6.3.2 Motion of Lines in Three Views.- 6.3.3 Conclusion on Lines and Points Motion.- 6.4 Implementation and Experimental Results.- 6.4.1 Implementation of the Motion Module.- 6.4.2 Experimental Results.- 6.5 Conclusion.- 6.6 Application to the Planar Case.- 7. Conclusion.- References.

O autorze

The author Thierry Viéville is a Researcher at the National Research Institute in Computer Science and Control Theory (INRIA) in France where he works in the Computer Vision and Robotics group headed by Olivier Faugeras. He is also an Associate Professor of the Nice University where he teaches active and robotic vision. Prof. Viéville participates in different European projects on computer vision.

Blurb

A Few Steps Towards 3D Active Vision describes several modules that can be used to perform 3D vision. A specific problem in the field of active vision is analyzed, namely how suitable is it to explicitly use 3D visual cues in a reactive visual task? The author has collected a set of studies on this subject and has used these experimental and theoretical developments to propose a synthetic view on the problem, completed by some specific experiments. With this book scientists and graduate students will have a complete set of methods, algorithms, and experiments to introduce 3D visual cues in active visual perception mechanisms such as autocalibration of visual sensors on robotic heads and mobile robots. Analogies with biological visual systems provide an easy introduction to this subject.