CAREER: Smart Cameras Getting Smarter: Detecting High-level Events Across Battery-powered Wireless Embedded Smart Cameras

视觉监控广泛应用于军事、商业和公共交通场景中。据估计，美国有3000万个监控摄像头，每周捕捉40亿个小时的镜头。由于观看多个视频源的人类操作员的注意力随着时间的推移而降低，因此由这些摄像机生成的大量视频数据需要自动检测语义上的高级别事件。此外，要求摄像头接入电源插座和有线链路会妨碍系统在摄像头数量和位置方面的灵活性，产生巨大的成本，并限制可能的应用和移动性。嵌入式智能摄像机允许我们部署许多空间分布的摄像机，通过无线链路相互连接。智能相机在单个嵌入式平台上集传感、处理和通信于一体。然而，它的能量和处理能力非常有限。该项目的目标是构建一个电池供电的、自适应的无线嵌入式智能摄像机系统，用于检测语义高层事件，以可扩展和节能的方式跨越多个重叠或非重叠的摄像机视图，并消除对有线链路的依赖。将开发资源感知和分布式目标跟踪和事件检测算法，以及自适应决策方法，以降低能耗，提高相机的电池寿命。预计这些结果将产生重要的积极影响，因为它们将通过同时解决隐私问题从根本上改变视频监控和事件检测解决方案。这项技术可用于跨学科和广泛领域，包括下一代网络物理系统、军事和商业应用、交通分析、医疗保健和野生动物监测。通过在当地儿童S动物园和博物馆部署这一项目，计划对年幼的儿童进行外联。

Visual surveillance is widely used in military and commercial applications, and public transportation scenarios. There are an estimated 30 million surveillance cameras in the U.S. capturing 4 billion hours of footage a week. The large amounts of video data generated by these cameras necessitate automatic detection of semantically high-level events, since the attention of a human operator watching multiple video feeds decreases over time. In addition, requiring cameras to have access to electrical outlets and to have wired links hinders system flexibility in terms of the number and placement of cameras, incurs significant costs, and limits possible applications and mobility. Embedded smart cameras allow us to deploy many spatially-distributed cameras interconnected by wireless links. A smart camera combines sensing, processing, and communication on a single embedded platform. Yet, it has very limited energy and processing power. The objective in this project is to build a battery-powered, self-adapting, wireless embedded smart camera system for the detection of semantically high-level events, which can span multiple overlapping or non-overlapping camera views, in a scalable and energy-efficient manner, and remove the dependence on wired links. Resource-aware and distributed object tracking and event detection algorithms, and self-adapting decision methodologies will be developed to decrease the energy consumption, and increase the battery-life of cameras. The outcomes are expected to have important positive impact, because they will fundamentally transform video surveillance and event detection solutions by addressing the privacy issues simultaneously. This technology can be used across disciplines and in wide-ranging areas, including next-generation cyber-physical systems, military and commercial applications, traffic analysis, health care and wildlife monitoring. Outreach to younger children is planned through the deployment of this project at the local Children?s Zoo and Museum.