Goal driven visual object identification and tracking system for a heterogeneous sensor network

智能空间是一个物理环境，它充满了感官、计算和通信能力，以及将它们集成在一起的软件。它将预见住户的需要，并在他们需要时提供服务。这些空间在医疗保健、制造、教育、环境监测、指挥中心和安全等不同领域都有应用。使这种系统成为现实所需的一个关键组成部分是识别、定位和跟踪该空间中的物体（包括有生命的和无生命的）的能力。这种能力还为确定智能空间的居住者之间的活动模式铺平了道路。在本提案中，我们讨论了软件框架的开发以及一组用于传感器数据收集和分析的工具，这些工具将实现用于传感器网络的此构建块。该系统中的每个工具都封装了一个算法，该算法提供了特定的功能，如人脸检测、语音识别和物体定位等。更重要的是，每个工具本身也将包含使用算法所需的知识。这些知识包括关于输入数据、算法参数、输出格式的信息，以及关于它所能完成的任务以及它如何很好地完成这项任务的信息。我们在本提案中的主要贡献是开发软件基础设施，允许专家工具社会完成给定的任务，开发单个工具的模板，开发单个工具的知识表示机制和词汇表，以及允许某些工具扩展其功能的基于示例的学习方法。我们计划调查的一个关键领域是智能空间的安全和隐私方面，这一点现在受到了相当大的关注。我们建议为智能空间制定一套安全和隐私要求，并开发支持这些要求的算法和机制。

A smart space is a physical environment that is saturated with sensory, computing and communication capabilities together with the software that integrates them. It will anticipate the needs of its occupants and will provide services as and when they are needed. Such spaces have applications in diverse areas such as healthcare, manufacturing, education, environment monitoring, command centers and security. A key building block that is needed to make such a system a reality is the capability to identify, locate and track objects (both animate and inanimate) in this space. Such a capability also paves way for determining patterns of activity among the occupants of a smart space. In this proposal, we discuss the development of a software framework together with a set of tools for sensory data collection and analysis that will implement this building block for use in a sensor network.Each tool in this system will encapsulate an algorithm that provides a particular functionality such as face detection, speech recognition and object location etc. More importantly, each tool will also incorporate the knowledge required to use the algorithm within itself. Such knowledge includes information about input data, algorithm parameters, output formats as well as information about the task it can accomplish and how well it can accomplish this task.Our primary contribution in this proposal is the development of the software infrastructure that allows a society of expert tools to accomplish a given task, development of  templates for individual tools, development of knowledge representation mechanisms and vocabularies for individual tools as well as example based learning methods that allow certain tools to expand their capabilities.A critical area that we plan to investigate is the aspects of security and privacy in a smart space, which are now gaining considerable attention. We propose to develop a set of security and privacy requirements for a smart space as well as develop the algorithms and mechanisms to support these requirements.