SBIR Phase I: EI:I2: PANOPTES: A Sytem for Analyzing Big Unstructured Multimedia Data Collections with Applications in Law Enforcement, Manufacturing, and Retail

该小型企业创新研究(SBIR)第一阶段项目将开发一个用于索引、搜索和可视化大型非结构化多媒体数据(照片、视频和录音)集合并将其商业化的系统。在发生恐怖袭击等意外灾难后，执法部门被来自许多不同来源的大量多媒体资产淹没，从安全摄像头到目击者智能手机。对于调查人员来说，快速可靠地识别这些大型语料库中潜在的相关镜头至关重要，以帮助他们进行调查。该项目将开发先进的软件来满足这一需求。该项目的广泛影响/商业潜力源于其在多个行业的适用性，包括执法、国防、制造和商业零售。据估计，美国这些市场的总规模约为6万亿美元。该项目解决了对组织大量非结构化多媒体数据的自动化工具的需求，预计随着智能手机的增长、低成本记录设备的普及以及监控和安全视频系统的普及，这些数据的数量和大小将会增加。目前搜索这些类型的数据收集的方法将被证明是不够的，因为它依赖于旨在审查少量素材的工具，这些素材通常是由少数设备在已知的时间点和空间点捕获的。

This Small Business Innovation Research (SBIR) Phase I project will develop and commercialize a system for indexing, searching, and visualizing large unstructured collections of multimedia data (photos, videos, and audio recordings). In the aftermath of an unexpected disaster such as a terrorist attack, law enforcement is overwhelmed with a vast amount of multimedia assets from many different sources ranging from security cameras to witness smartphones. It is critically important for investigators to quickly and reliably identify potentially relevant footage in these large corpora in order to aid their investigation. This project will develop advanced software to address this need. The broader impact/commercial potential of this project results from its applicability across a number of industries, including law enforcement, defense, manufacturing, and commercial retail. The combined size of these markets in the US is estimated to be roughly $6 trillion. This project addresses the need for automated tools to organize large unstructured collections of multimedia data that are expected to increase in number and size in parallel with the growth in smartphones, wider availability of low-cost recording equipment, and the prevalence of surveillance and security video systems. The current approach for searching these types of data collections will prove inadequate, since it relies on tools that are designed for reviewing a small amount of footage, normally captured by a small number of devices at known points in time and space.