ESPRIT COLLABORATIVE: Control of Perception in Active Vision

小行星9303980 这是一个为期三年的连续奖励的第一年，以支持研究人员在四个美国机构（宾夕法尼亚大学，罗切斯特大学，马里兰州和马萨诸塞州大学）与他们的欧洲同事从事ESPRIT资助的研究项目题为“视觉作为过程“的合作。 合作的目标是开发一个实用的理论，感知控制主动视觉，同时汇集专业知识，避免不必要的重复工作。 广泛的重点领域是开发用于集成视觉/机器人系统的新型物理和计算机硬件、概念框架和软件工具，其中包括主动控制的双目摄像头和用于影响外部世界的各种致动器。 该合作涵盖四个子领域：主动视觉中的感知控制，动态场景的解释，鲁棒的图像和场景测量，以及动态凝视控制和图像形成。 将为每一个分领域建立一个信息交流邮件名单，举行科学会议，并组织技术和相关实验。 为期两到三周的伙伴间访问将促进在具体专题上的密切合作。 美国博士作为交换协议的一部分，学生将在EEC实验室呆上一段时间。 双方将受益于一个演示的实时主动视觉系统，以及一个集成和控制实验的通用框架的开发。 这样的系统将扩大同类问题的实验种类和不同领域的应用范围。 将通过计划出版一本关于理论和实验结果的书、期刊论文和在会议上的演讲，使研究结果广泛传播。 ***

9303980 Bajcsy This is the first year of a three-year continuing award to support collaboration of researchers in four U.S. institutions (University of Pennsylvania, University of Rochester, University of Maryland, and University of Massachusetts) with their European colleagues engaged in the ESPRIT-funded research project entitled "Vision-As- Process ". The collaborative objective is to develop a practical theory of perception control in active vision, while pooling expertise and avoiding unnecessary duplication of effort. The broad areas of concentration are development of novel physical and computer hardware, conceptual frameworks, and software tools for integrated vision/robotic systems, which include an actively controlled binocular camera head and various actuators for affecting the external world. The collaboration covers four subareas: Control of perception in active vision, Interpretation of dynamic scenes, Robust image and scene measures, and Dynamic gaze control and image formation. For each subarea a mailing list for information exchange will be created, scientific meetings held, and techniques and associated experiments organized. Two to three week partner-to-partner visits will facilitate intense collaboration on the specific topics. US Ph.D. students will spend time in EEC laboratories as part of an exchange agreement. Both sides will benefit from a demonstrated real-time active vision system, and development of a common framework for integration and control experiments. Such a system will enlarge the variety of experiments within the same class of problems and the range of applications in different domains. The results will be made widely accessible, through planned publication of a book on the theoretical and experimental results, journal papers, and presentations at conferences. ***