I-Corps: Automated Spatiotemporal Intelligence Operations for Asset Integrity Management

University Of New Mexico, Albuquerque NM

Investigators

Abstract

The broader impact/commercial potential of this I-Corps project includes the enabling of wide scale exploitation of the growing volume of airborne image data from unmanned aerial systems (UAS) to monitor the environment in near real-time. Persistent and tactical surveillance of infrastructure assets (e.g., critical infrastructure, roads, pipelines), military bases, agricultural fields, boarders, or other extensive assets requiring routine monitoring for tactical decision making is becoming cost feasible with the introduction of UAS, but the volume of data collected cannot be exploited using traditional, largely manual, methods. Automated processing and interpretation of large volumes of airborne imagery in near real-time will enable improved decision making and, subsequent, cost reductions and improved performance for a range of industries and agencies. The operations of infrastructure management, disaster response, security, intelligence, and agriculture are amongst the expected beneficiaries. This I-Corps project explores the commercialization potential of a platform from near real-time analysis and exploitation of airborne image data. Research exploring the development of an airborne system for monitoring critical infrastructure during the response phase on natural disasters resulted in the development of an analytical model for repeat station imaging and refinement of a conceptual model for the design of time-sensitive remote sensing systems that collectively permit the design and implementation of automated change detection and monitoring systems from airborne imaging. This spatial analytics platform automates 3D scene reconstruction, and uses artificial intelligence and machine learning techniques to convert digital photos into a cataloged and indexed spatial intelligence database of changes over time. 3D/4D object classifiers are developed to extract complex features that 2D imagery is unable to represent. This method trains neural networks on volumetric data and multi-temporal spatial data to facilitate the extraction and identification of features and how they have changed. Object identification and characterization will provide the capability to semantically describe changes 3D and 4D space.