Vincze M., Hager G.D. (eds.) Robust Vision for Vision-Based Control of Motion

Издательство SPIE Optical Engineering Press / IEEE Press, 2000, -247 pp.Understanding how to make a "seeing robot" - this was the goal of the workshop Robust Vision for Vision-Based Control of Motion held at the 1998 IEEE International Conference on Robotics and Automation in Leuven, Belgium. As far-fetched as this goal might seem, the continuing development of computers, robotic hardware, and techniques for machine vision suggests that it inevitably will be reached.
It is clear that, given a sufficiently structured environment, today' s robots can perform complex tasks with astonishing precision and speed. Hence it would seem that, from the standpoint of mechanics and control, robotics research has already advanced to the point where it could move out of the structured environment of the factory floor and into the "real" world. Why has this not happened? The limiting factor, simply stated, is the ability to perceive and react in such complex and unpredictable surroundings. In particular, vision - as the most versatile sense of perception - is required to aid the navigation, grasping, placing, steering, and the control of motion of a machine, be it a robot arm, a vehicle, or any other autonomous mechanism. Thus, it may be argued that robust vision is the key bottleneck to the development of autonomous robotic mechanisms.
For this reason, the use of visual sensing to motion control has drawn increasing interest over the last several years. The level of interest is documented by a series of Workshops and Tutorials at ICRA and other conferences as well as a marked growth in the number of sessions devoted to the topic at recent conferences. As a result, there have been many advances in the science and practice of vision-based control of motion. Yet, most laboratory vision-based control systems are constrained to use markers, special high contrast objects and background surfaces, or detailed models and good initialization - customizations often not possible in real applications.
This book, as well as the workshop from which it is derived, was motivated by the observation that fast, cheap, yet robust and reliable vision is still a major roadblock to industrial and service applications of vision-based control. Our objective is to present the most recent work dealing with this essential problem in visual feedback: robustness. As such, the collection contains a series of articles spanning a large range of issues including hardware design, system architecture, control, sensor data fusion, object modeling, and visual tracking. The 14 papers represent major research groups in Europe, America, and Japan. The following summarizes the contributions.Cue Integration for Manipulation
Spatiallyadaptive Filtering in a Modelbased Machine Vision Approach to Robust Workpiece Tracking
Incremental Focus of Attention: a Layered Approach to Robust Vision and Control
Integrated Object Models for Robust Visual Tracking
Robust Visual Tracking by Integrating Various Cues
Two-Dimensional Model-Based Tracking of Complex Shapes for Visual Servoing Tasks
Interaction of Perception and Control for Indoor Exploration
Real-Time Image Processing for Imagebased Visual Servoing
Proventechniques for Robust Visual Servo Control
Global Signatures for Robot Control and Reconstruction
Using Foveated Vision for Robust Object Tracking: Three-Dimensional Horopter Analysis
Evaluation of the Robustness of Visual Behaviors Through Performance Characterization
Robust Image Processing and Positionbased Visual Servoing
Vision-Based Objective Selection for Robust Ballistic Manipulation
Vision-Based Autonomous Helicopter Research at Carnegie Mellon Robotics Institute (1991-1998)