Multimodal Integrated Remote Sensing for Urban Environments

这项研究将开发新的计算方法来处理遥感数据，以监测城市环境。具体而言，开发的方法将着眼于解决处理多模态数据时出现的挑战，以便利用每种观测模式的互补优势。对这些数据的分析提出了许多与计算机视觉领域中的经典问题不同的问题，其中输入数据最常使用捕获可见波长的传感器来获取，并且通常本质上是欧几里得的。虽然处理固有的非欧几里德数据存在许多挑战，但本研究旨在考虑几何深度学习在多模态遥感数据分析中令人兴奋的发展的优点。以下EPSRC研究领域已被确定为本研究将参与的主题：人工智能技术;建筑环境;数据信号处理;图像和视觉计算;基础设施和城市系统;统计和应用概率;结构工程。

This research will develop novel computational methods to process remotely sensed data to monitor urban environments. Specifically, the developed methods will look to tackle challenges that arise when processing multimodal data such that the complementary strengths of each observation mode can be harnessed. The analysis of such data presents many different problems to classic problems in the field of computer vision, where input data is most commonly acquired using sensors which capture visible wavelengths and is generally Euclidean in nature. While working with data which is inherently non-euclidean presents many challenges, this research aims to consider, among other techniques, the merits of exciting developments in geometric deep learning for the analysis of multimodal, remotely-sensed data. The following EPSRC Research Areas have been identified as topics which this research will engage with: Artificial intelligence technologies; Built environment; Data signal processing; Image and vision computing; Infrastructure and urban systems; Statistics and applied probability; Structural engineering.