Visual Perception in Ad Hoc Networks of Smart Cameras

我设想的未来是，由固定式固定相机、配备相机的手持移动设备和安装在自主平台上的相机组成的成像网络将收集和分析大量图像和视频，以支持基础设施管理、生态监测、独立生活的智能空间、灾难准备和响应、交通系统、个性化照片代理等领域的各种应用。这些网络将比现有的“多相机系统”大一个数量级，许多科学和技术挑战仍然存在，需要进行大量原创性研究才能实现这一愿景。这项提案寻求资金，以加快开发能够无人值守运行的下一代智能相机网络，以支持城市传感、智能环境以及场景分析和建模。在接下来的五年里，我们的目标是：1)发展智能摄像机自组织网络中协作感知的科学理论；2)发展总结这些网络捕获的视频的理论和方法；3)探索交叉视点活动识别的科学和理论基础。这项提案中概述的工作将带来新的科学发现、数学理论和智能相机网络中的视觉感知工具。这项工作还将使无数智能相机网络应用成为可能。该提案还为独特的跨学科高素质人才(HQP)培训提供了令人兴奋的机会。这些HQP不仅将在计算机视觉、机器学习、相机网络和系统建设领域发展深厚的技术和科学知识，而且还将学习人际交往和口头和书面沟通技能。我预计，在当前以技术为重点的就业市场上，这些HQP的需求将会很大。这些HQP也将有机会进入学术界，如果这是利益所在的话。该项目的成果将直接惠及本地和国际摄像技术、视频分析和数字娱乐产业，刺激经济增长。根据这项提议开展的工作将遵循大学关于公平、多样性和包容性的指导方针。

I envision a future where imaging networks comprising stationary fixed cameras, hand-held camera-equipped mobile devices, and cameras mounted on autonomous platforms will collect and analyze large sums of imagery and video to support a variety of applications in areas such as infrastructure management, ecological monitoring, smart spaces for independent living, disaster readiness and response, transportation systems, personalized photo agents, etc. These networks will be an order of magnitude larger than the existing "multi-camera systems," and many scientific and technical challenges remain and much original research is needed to realize this vision. This proposal seeks funds to accelerate the development of the next generation of smart camera networks capable of unattended operation in support of urban sensing, smart environments, and scene analysis and modelling. Over the next five years, we aim to: 1) develop scientific theory for collaborative sensing in ad hoc networks of smart cameras; 2) develop theory and methods for summarizing videos captured by these networks; and 3) explore scientific and theoretical foundations for cross-view activity recognition. Work outlined in this proposal will lead to new scientific discoveries, mathematical theories, and tools for visual perception in smart camera networks. The work will also enable a myriad of smart camera networks applications. The proposal also provides exciting opportunities for unique interdisciplinary Highly Qualified Personnel (HQP) training. These HQPs will not only develop deep technical and scientific knowledge in the areas of computer vision, machine learning, camera networks, and system building, they will also learn interpersonal and oral and written communication skills. I expect that these HQPs will be in high demand in the current technology-focused job market. These HQPs will also have the opportunity to go into academia if that is where there interests lie. The outcomes of this program will directly benefit both local and international camera technology, video analytics, and digital entertainment industries, spurring economic growth. The work carried out under this proposal will adhere to the university guidelines around equity, diversity and inclusivity.