STTR Phase I: Rapid and Efficient Scene Modeling for Law Enforcement and Disaster Response

这项小型企业技术转移研究（STTR）第一阶段项目将生产能够准确、快速、低成本地构建场景三维（3D）模型的技术。该技术利用了计算机视觉的最新发展，增强了它们在不同环境条件下的操作能力，同时提高了模型构建的速度，从而实现了实际应用。确定的关键系统设计要求包括：1)能够快速生成完整、准确的高保真模型；2)对不同环境条件的鲁棒性；3)非专业人员无需经过广泛培训即可轻松使用；4)成本低。提出的研究目标包括：改进最先进的算法，以实现图像的鲁棒和高效3D建模；开发了一种使用不同传感器模式进行密集3D重建的方法；一种传感器视点规划算法的研制创建识别缺失模型数据的技术；以及用于将来自不同来源的数据合并为环境的连贯3D模型的系统的开发。工作计划包括对系统的全面评估。S准确、完整地重建场景的能力，以及使用低成本车辆部署的传感器重建场景的好处。这个项目更广泛的影响/商业潜力将是场景文档中3D建模的广泛使用。目前可用于此应用程序的系统价格昂贵、速度慢、体积大，并且需要经过特殊培训才能使用。这项工作产生的技术将使系统的生产和商业化更快、更经济、更容易被非专业人员使用，从而简化大量执法人员、检察官、保险公司和其他政府机构的采用。例如，在交通领域，当事故发生时，快速的场景重建对于恢复交通流量至关重要。在繁忙的高速公路上，清理事故现场的交通拥堵可能会以每分钟一英里的速度增加。目前，这一规定将事故的完整记录仅限于发生死亡的情况。提出的技术将允许记录更多的事件。此外，预计将大大减少确定事故原因所涉及的时间和复杂性。这种减少的潜在长期利益包括更迅速地制定规章和政策，以处理交通事故的长期原因，这将加快道路安全机制的执行。

This Small Business Technology Transfer Research (STTR) Phase I project will produce technology capable of constructing three-dimensional (3D) models of scenes accurately, rapidly, and at low cost. The technology leverages recent developments in computer vision and enhances them for operation under varying ambient conditions while increasing speed of model construction, thus enabling practical applications. Key system design requirements identified include: 1) ability to produce complete and accurate high fidelity models quickly; 2) robustness to varying ambient conditions; 3) ease of use by a non-specialist without extensive training; and 4) low cost. The proposed research objectives include: improvements to state-of-the-art algorithms for robust and efficient 3D modeling from images; the development of an approach for dense 3D reconstruction using different sensor modalities; the development of a sensor view point planning algorithm; the creation of techniques for the identification of missing model data; and the development of a system for merging data from different sources into a coherent 3D model of the environment. The work plan includes a thorough assessment of the system?s ability to reconstruct a scene accurately and completely, as well as the benefit of reconstructing scenes using sensors deployed by low-cost vehicles. The broader impact/commercial potential of this project will be the widespread use of 3D modeling for scene documentation. Systems currently available for this application are expensive, slow, bulky, and their use requires special training. The technology resulting from this work will enable the production and commercialization of systems that are faster, more affordable, and easier to use by non-experts, thereby simplifying their adoption by a larger number of law enforcers, prosecutors, insurers, and other government agencies. In the transportation domain, for example, a speedy scene reconstruction is essential to restore the flow of traffic when accidents happen. On busy highways, traffic backup can grow at a rate of up to a mile per minute of delay in clearing the accident site. This currently limits the complete documentation of accidents to only those cases where fatalities occur. The technology proposed will allow the documentation of larger number of incidents. Additionally, a great reduction in the time and complexity involved in determining the cause of accidents is expected. The potential long-term benefits of this reduction include a more rapid evolution of regulation and policy to deal with chronic causes of traffic accidents, which will expedite the implementation of road safety mechanisms.