A data-driven platform for multimodal positioning and tracking in indoor environments

定位和跟踪人在室内环境中的活动对于基于位置的服务(LBS)至关重要，例如室内路线查找、基于邻近的广告、劳动力管理、地理围栏、欺诈检测等。现有的室内定位系统(IPS)在成本、准确性、可用性、理解用户与环境之间的交互以及在无设备环境下工作等方面存在不足。该项目建立在我们在IPS、邻近技术和数据融合方面的丰富经验基础上，以开发可用于各种市场(例如，零售、医疗、仓库等)的亚米精度多模式IPS解决方案。我们设想的平台由以下组件组成：(1)监控感兴趣区域的分布式相机网络，以及个人设备上的移动传感器；(2)实时视觉数据处理和定位，以识别和跟踪对象以及人的活动；以及(3)数据准备和分析引擎，用于清理和整理数据以进行分析。在本提案中，我们将解决实时、多模流数据处理带来的算法和计算挑战，以开发准确可靠的室内定位平台。*

Locating and tracking people's activities in indoor environments are crucial to location-based services (LBS) such as indoor wayfinding, proximity-based advertisement, workforce management, geo-fencing, fraud detection, etc. Existing indoor position systems (IPS) fall short in cost, accuracy, usability, understanding of interactions between users and environment, and working in device-free settings. The project builds upon our extensive experience in IPS, proximity technologies, and data fusion to develop a multimodal IPS solution at sub-meter accuracy that can be used in a variety of markets (e.g, retail, health, warehouse, etc.). Our envisioned platform consists of the following components: (1) a distributed camera network that monitors the areas of interest, along with mobile sensors on personal devices; (2) real-time vision data processing and localization to identify and track objects as well as people's activities; and (3) a data preparation and analysis engine that cleans and curates the data for analytics. In this proposal, we will address both algorithmic and computational challenges posed by processing in real-time, multi-modal streaming data to develop an accurate and reliable indoor localization platform.**********