CSR: Medium: Collaborative Research: Self-Coordination in Cooperative Smart Camera Networks Incorporating System-On-Chip Reconfiguration

随着技术的进步和相机系统价格的下降，我们生活中的相机数量和相机系统的规模不断增加，创造了新的机会和应用。除了个人用途外，摄像机还广泛用于军事、公共和商业应用，用于监视和统计数据收集。据估计，美国有3000万个监控摄像头，每周拍摄40亿小时的镜头。除了用于监控目的的传统摄像机外，谷歌Glass等项目正在推动具有本地处理和通信能力的微型和低成本摄像机的发展。对于未来的摄像系统，局部智能和组件之间的自主协作将提供解决更复杂任务的能力，这需要一个统一的视角来同时解决硬件/软件协同设计、实时操作、高精度以及运行时的自协调和自适应的挑战。本项目为一组协作嵌入式智能摄像头的设计、部署和自协调提供了一种整体的、新颖的方法，目标是以最高的精度和最小的延迟监控大面积。其中一个目标是为嵌入式智能相机设计综合方法和计算基础设施，允许硬件重构和系统地在硬件和软件之间实时交换任务。另一个目标是开发自配置方法，以自主适应系统行为，并以最佳方式处理运行时环境变化，包括节点故障。这项研究有望通过为设计和部署此类系统提供系统方法，并在实验室和校园规模上测试新方法，从而开发出新的实时、全自动、协作和高度精确的相机系统。潜在的应用包括智能监控系统、基于多摄像头的驾驶辅助系统、养老院的辅助、基于3D重建的生产线质量控制以及远程手术。该项目还将研究与两个机构的本科和研究生课程相结合，并通过阿肯色大学工程职业意识项目、阿肯色路易斯斯托克斯少数民族参与联盟和乔治华盛顿卡弗项目，以及锡拉丘兹大学的WiSE项目，为增加代表性不足群体的参与做出了贡献。来自代表性不足群体的学生将被招募并参与协作多摄像机网络的设计、实施和部署。

The number of cameras in our lives and the scale of camera systems are continuously increasing as technological advances and falling prices in camera systems create new opportunities and applications. In addition to personal uses, cameras are widely employed in military, public and commercial applications for surveillance and statistics gathering. There are an estimated 30 million surveillance cameras in the U.S. capturing 4 billion hours of footage a week. Besides the traditional use of cameras for surveillance purposes, projects such as Google Glass are driving the development of miniature and low-cost cameras with local processing and communication capabilities. For future camera systems, local intelligence and autonomous collaboration among components will provide the capability to solve more complex tasks, which requires a unifying perspective to simultaneously address the challenges of hardware/software co-design, real-time operation, high accuracy and self-coordination and self-adaptation in run-time.This project provides a holistic and novel approach for the design, deployment and self-coordination of a set of collaborative embedded smart cameras, with the goal of monitoring large areas with the highest accuracy and smallest latency. One objective is designing synthesis approaches and computing infrastructure for the embedded smart cameras that allow hardware restructuring and systematic swapping of tasks between hardware and software on-the-fly. Another objective is to develop self-configuration approaches to autonomously adapt system behavior and optimally deal with run-time environmental changes, including node failures.This research is expected to enable development of new real-time, fully automated, collaborative and highly accurate camera systems by providing a systematic approach for the design and deployment of such systems, and testing new methods at laboratory and campus scales. Potential applications include smart surveillance systems, multi-camera-based driver assistance systems, assistance in nursing homes, quality control on production lines based on 3D reconstruction, and remote surgery. The project also integrates research with the undergraduate and graduate programs of two institutions and contributes towards increasing the involvement of under-represented groups through the University of Arkansas Engineering Career Awareness Program, Arkansas Louis Stokes Alliance for Minority Participation and George Washington Carver Project, and the WiSE program at Syracuse University. Students from under-represented groups are to be recruited and involved in the design, implementation, and deployment of collaborative multi-camera networks.