ITR: Distributed Smart Cameras: Algorithms, Architectures, and Synthesis

我们建议通过对分布式算法、嵌入式体系结构和软件合成技术的综合探索，为分布式智能摄像机网络开发新的技术和方法。我们将开发旨在处理现代视频处理算法的新架构和工具。这些算法将利用分布式体系结构来实现高效的实现。我们将研究从监控和生物识别应用中提取的一系列复杂的智能相机算法。具体地说，我们将研究分布式摄像机自校准、使用基于图像的视觉外壳的视图合成以及使用人脸和步态进行人体识别的算法的有效实现。分布式摄像机网络是一种新兴的使能技术，适用于广泛的重要信息技术应用，包括监控、生物识别和智能会议室。由于可以从多个方向访问场景，这样的网络比单摄像头系统有可能获得完整的视图。这有助于提高算法的健壮性和通用性。然而，这种分布式的性质，再加上与实时视频处理相关的内在挑战，使得开发有效的算法、体系结构和软件变得非常复杂。需要一个包括视频、设计工具和嵌入式系统体系结构的综合研究计划来了解如何最大限度地利用新一代智能摄像机硬件。

We propose to develop novel techniques and methodologies for distributed smart camera networks through an integrated exploration of distributed algorithms, embedded architectures, and software synthesis techniques. We will develop new architectures and tools that are designed to tackle modern video processing algorithms. The algorithms will leverage distributed architectures for efficient implementations. We will investigate a series of complex smart camera algorithms drawn from surveillance and biometrics applications. Specifically, we will be investigating efficient implementations of algorithms for self-calibration of distributed cameras, view-synthesis using image-based visual hulls, and human identification using face and gait. Networks of distributed cameras are an emerging enabling technology for a broad range of important information technology applications, including surveillance, biometrics, and smart conference rooms. By having acces to scenes from multiple directions, such networks have the potential for complete views than single-camera systems. This contributes to increased robustness and generality of algorithms. However, this distributed nature, coupled with the inherent challenges associated with real-time video processing, greatly complicates the devlopment of effective algorithms, architectures and software.An integrated research program inclduing video, design tools, and embedded system architecture is required to understand how new generation ofsmart camera hardware can be best utilized.