Integration of Artificial Intelligence into the US Intelligence System: An Analysis of the Current Situation of Artificial Intelligence Enhancing Geospatial Intelligence

The United States Imagery and Geospatial System believes that artificial intelligence can have an impact on human-intensive task analysis, target identification, task automation, geospatial intelligence agent training, and methodological research to advance intelligence technology. The National Imagery and Mapping Agency (NIMA), the predecessor of the National Geospatial-Intelligence Agency

(NGA), began researching AI solutions in imagery intelligence very early on. NIMA research shows that establishing accurate and timely situational awareness is critical during mission execution. Artificial intelligence can integrate the available information of the target area into a situation map, generate 2D or 3D images and realize the construction of a dynamic common operating picture (COP). In addition, many artificial intelligence projects of the Defense Advanced Research Projects Agency (DARPA) are also closely related to automatic target recognition and geospatial target analysis. DARPA released the Moving Target Recognition program in July 2020. Its purpose is to solve the problem of using synthetic aperture radar (SAR) to detect, image and geolocate ground moving targets by developing processing algorithms and collection technologies.

Change detection is another great application of artificial intelligence in geospatial intelligence. Traditionally, interpreting and analyzing satellite images of global targets sent back every day and generating intelligence reports is the most basic daily work of NGA image analysts, which usually requires a lot of human resources. But by learning from historical patterns, current trends and dynamic factors such as weather and adversarial behavior, AI can improve the ability to predict target behavior, helping to determine the most appropriate methods and locations to deploy assets for surveillance. In other words, for fixed targets such as military and industrial facilities in the target country, by improving the artificial intelligence change detection algorithm, the machine can independently complete most of the monitoring and screening work, helping image analysts prioritize high-value images, find target objects, identify abnormal activities, monitor changes in target facilities, and finally realize automatic early warning based on images.

At present, the software used to assist image analysis has developed from correction, storage and labeling functions to intelligent interpretation software that can quickly identify and detect a large number of images. In the face of image data from heterogeneous and multi-platform, intelligence agencies need to invest a lot of resources in processing and analysis. Today, however, state-of-the-art automated image analysis software can combine disparate datasets from different providers, including attributes such as different resolutions, coordinate systems, and multiple spectral bands. However, at the moment when the development of artificial intelligence technology is not yet perfect, the positioning of artificial intelligence in geospatial intelligence is still controversial. There is no exact way to solve the problem of discrepancy between artificial intelligence and analyst analysis results. Therefore, deepening data mining, automatic identification and data visualization and other auxiliary analysis intelligence enhancement means is the main development direction of artificial intelligence in the field of geospatial intelligence.