ATD: Threat Detection Problems in Precision Agriculture and Satellite Imaging

该项目通过利用对精准农业和卫星成像中相关未解决问题的研究，探索可能的威胁检测技术。其中一些问题目前在实验室环境中得到了解决，但在遥感的实际环境中还没有令人满意的解决方案。除了安全方面的好处之外，这项研究还有可能通过只在需要的时间和地点使用所需的投入物来节省种植者的资金并减少农业对环境的影响。该项目对卫星成像和测量的贡献也将加强环境监测。有效的遥感将使我们能够从最少的数据中准确地探测到威胁，同时延长扫描仪的使用寿命并降低相关成本。这个项目的目的是发展必要的数据清理和处理的关键方法，以确保航空和卫星图像的有意义的结果。例如，研究人员将探索一种谨慎而新颖的方法来去除卫星图像中的云。一种新的方法是必要的，因为植被指数的测量有助于识别对作物的各种威胁，在云带来的阴影的存在下变得不可靠。研究者和合作者还将解决图像拼接的技术问题，或解决农业领域的大型难题。这里的困难在于补丁彼此非常相似，有时很少有重叠。其他需要研究的技术问题是超分辨率和不均匀环境下的上漆。一旦这些技术问题得到解决，研究人员计划适应并进一步发展威胁检测技术，特别是多元变化点检测技术。强大的农业威胁检测也将有助于优化作物产量，同时最大限度地减少资源。因此，在这些领域的威胁检测能力的进步，广泛地影响社会，甚至超出其内在的用途。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This project explores possible techniques for threat detection by leveraging research into related unsolved problems in precision agriculture and satellite imaging. Some of these problems are currently solved in a laboratory setting but do not have satisfying solutions in the practical setting of remote sensing. In addition to security benefits, the research has the potential to save growers money and reduce the environmental impact of agriculture by applying only the inputs that are needed only when and where they are needed. The project's contributions to satellite imaging and measurement would also enhance environmental monitoring. Efficient remote sensing would enable accurate threat detection from minimal amounts of data, while increasing the life span of scanners and reducing associated costs.This project aims to develop crucial methods for the data cleaning and processing necessary to ensure meaningful results in aerial and satellite imagery. For example, the investigators will explore a careful and novel approach to cloud removal in satellite images. A new approach is necessary since the measurement of vegetation indices, which help identify various threats to crops, becomes unreliable in the presence of shadows introduced by clouds. The investigator and collaborators will also address the technical problem of image stitching, or solving large puzzles, for agricultural fields. The difficulty here is that patches are very similar to each other and sometimes have very little overlap. Other technical problems to be studied are super-resolution and inpainting in non-uniform settings. Once these technical problems are solved, the investigators plan to adapt and further develop techniques for threat detection, in particular, techniques for multivariate change point detection. Robust threat detection in agriculture will also help to optimize crop yield while minimizing resources. Thus, advances in threat detection capabilities in these domains broadly impact society even beyond their intrinsic usefulness.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.