Information Fusion for Feature Extraction and the Development of Geospatial Information

TLDR

Geospatial intelligence for military missions requires timely, accurate extraction from imagery, but single-image methods are labor-intensive and often inadequate, so fusion-based approaches that combine multiple data sources are needed to overcome these limitations. The paper aims to develop automated tools that leverage fusion techniques to meet NGA’s geospatial intelligence needs. The authors analyze fusion-based methods that jointly process multiple data sources to enhance automated extraction. Recent studies show that automated data extraction tools add value, highlight research gaps, and suggest future directions.

Abstract

Abstract : Numerous military missions require timely, accurate geospatial intelligence, yet the extraction of geospatial information from imagery data is generally a labor-intensive process. Automated tools to assist analysts in this process would be of tremendous benefit for the production of geospatial intelligence at the National Geospatial- Intelligence Agency (NGA). Methods that rely on data from a single image, while useful, generally fall short of producing useable geospatial products. Fusion-based methods which jointly process and exploit information from multiple sources offer ways to address many of the limitations of single-source analysis. This paper discusses the needs for geospatial intelligence and examines the benefits of fusion to address these needs. Results from several recent studies demonstrate the value added from a number of tools for automated data extraction (ADE). Experience gleaned from these studies also indicates several areas where new research and development is needed. The paper concludes with a brief discussion of future directions.