SBIR Phase I: A Novel Platform for Automatic Geo-Tagging of Images

这个小型企业创新研究(SBIR)第一阶段项目旨在开发计算机软件，该软件可以自动识别数字照片中的建筑物、结构和地标，以及照片拍摄的准确位置。随着数字时代的到来，每天都有数以亿计的照片被分享到网上。这种由用户生成的内容的涌入，给组织和理解大型照片集带来了巨大的挑战，从一个家庭几年来拍摄的数千张照片，到每月在网上分享的数十亿张照片。例如，注释照片目前是一项繁重的任务；照片分享服务的用户没有准确的、自动的方法来标记他们照片中的地标和其他结构，而是依赖手动标记。利用最新发明的计算机视觉技术，该项目旨在将地点的照片自动转换为带有地理标记的用户参与门户网站故事板，其中包含丰富的像素级注释和超相关信息。这些创新利用了一个独特的3D模型数据库，提供了非常准确的位置和标签。该项目的目标包括开发世界范围的基于位置的图像内容识别算法、准确的注释方法和系统的应用程序编程接口，供应用程序开发商使用。该项目的广泛影响/商业潜力是巨大的；由于数字照片共享已成为一项主要的在线活动，市场将对识别照片并将其链接到超本地和高度相关的信息(包括广告)的能力产生相当大的市场兴趣，用于旅游和酒店市场部门以及当地小型企业。此外，这项技术将使社交媒体市场领域的公司能够通过互动和情景照片浏览来增加用户参与度。此外，拟议的技术可以将照片转换为移动电话市场部门的图像驱动平台，允许数字照片和在线地图之间的无缝过渡，从而极大地增强用户使用移动设备对新的物理环境的导航。最后，这项技术还可以通过确定人们在哪里以及为什么在感兴趣的地点拍照的聚合模式，为社交媒体监控中的公司提供关键分析。

This Small Business Innovation Research (SBIR) Phase I project aims to develop computer software that automatically recognizes buildings, structures, and landmarks in digital photographs, as well as the precise location where photos were taken. The advent of the digital age has led to hundreds of millions of photos being shared on the Web every day. This influx of user-generated content is leading to significant challenges in organizing and understanding large photo collections from the many thousands of photographs captured by a family over several years, to the billions of photos shared online each month. For instance, annotating photos is currently a burdensome task; users of photo-sharing services have no accurate, automated method for labeling landmarks and other structures present in their pictures, instead relying on manual tagging. Using newly invented computer vision technology, this project aims to automatically transform photographs of locations into geo-tagged user engagement portals storyboards rich with pixel-level annotations and hyper-relevant information. The innovations leverage a unique database of 3D models, yielding very accurate locations and tags. The project objectives include developing algorithms for world-scale location-based image content recognition, an accurate annotation method, and an application programming interface for the system for use by application developers.The broader impact/commercial potential of this project are significant; as digital photo-sharing has become a major online activity, there will be considerable market interest in the ability to identify and link photos to hyper-local and highly relevant information, including advertising, for the tourism and hospitality market sector, as well as for small local businesses. Additionally, the technology will enable companies in the social media market sector to increase user engagement through interactive and contextualized photo browsing. Furthermore, the proposed technology can transform photos into image-powered platform for the mobile phone market sector, allowing for a seamless transition between digital photos and online maps and thus greatly enhancing user navigation of new physical environments using mobile devices. Finally, the technology can also provide key analytics to companies in social media monitoring, by determining aggregate patterns in where and why people take photos at locations of interest.