NeTS-NOSS: Localized Computation and Network Path Formation to Enable Pervasive Video Sensing

传感器网络和视频数据流是两种成熟的科学观测和数据收集技术，它们的技术要求截然不同。前者的设计假设大量的单位和长期部署的设置与小数据生成率。后者可以在收集的数据中提供丰富的视觉细节，但需要大量的能源来维持数据记录、通信和存储。该项目的研究旨在使网络摄像机能够在传感器网络限制下运行：低能耗、远程无系绳设置和大空间测量尺度。这项工作的动机是与生态学家和生物学家合作，揭示了许多观察物种行为的机会，这些行为的特征是受人类存在干扰的事件，地点遥远，需要长时间的等待或需要大而详细的区域覆盖。在这些环境中进行检测、记录和流式传输以实现科学发现，可以扩展我们对环境的理解。该项目涉及开发一种新型低成本视频传感器，该传感器可以从环境中获取能量，并支持相机视场的空间和时间子采样。互补的研究重点包括本地化和协作的网络图像分析、数据压缩和网络路径形成的调查，以便将视频数据传输到外部观察者，同时最大限度地减少由多个流引起的争用。该项目包括在飞行员中使用由50个视频源组成的视频传感器现场进行演示，包括在波士顿大学萨金特营地观察林地动物，以及在与马萨诸塞大学野外站合作选择的沿海地点研究滨鸟和灰海豹。

Sensor networking and video data streaming represent two maturing technologies for scientific observation and data collection with very different technical requirements. The former is designed assuming large numbers of units and long-term deployment in settings with small data generation rates. The latter can provide rich visual detail in collected data but requires significant energy resources to sustain data recording, communication, and storage. The research in this project seeks to enable networked video cameras to operate within sensor network constraints: at low energy consumption, in remote un-tethered settings, and in large spatial measurement scales. The work is motivated by collaborations with ecologists and biologists that reveal many opportunities for the observation of species behavior that are characterized by events that are disturbed by human presence, are remotely sited, require long periods of waiting or require large, detailed area coverage. Detecting, recording, and streaming to enable scientific discovery in these settings can expand our understanding of the environment.The project involves the development of a novel low-cost video sensor that operates on energy harvested from the environment and supports spatial and temporal sub-sampling of the camera field of view. Complementary research thrusts include the investigation of localized and cooperative in-network image analysis, data compression, and network path formation to enable delivery of video data to an outside observer while minimizing contention caused by multiple streams. The project includes a demonstration using a video sensor field comprised of 50 video sources in pilots involving the observation of woodland animals at Boston University's Sargent Camp and in the study of shorebirds and grey seals at a coastal site selected in collaboration with the University of Massachusetts Field Station.