Language Enabled Raster Imagery Exploitation (LERIX)

Raster Maps, which represent a significant portion of our imagery assets library, have limited intelligence value in their native raster form. To unlock their value, these raster maps must undergo a complex transformation process that extracts metadata about the geo-coordinates, georegisters the scene, and extracts the texture information (features, roads, rivers) and text labels of features and locations. Text recognition is particularly challenging with respect to text segmentation (line separation), font and character code conversion, and language recognition. And with our nation at war, the text issues regarding Arabic-scripted language maps are particularly urgent. Our proposal - Language Enabled Raster Imagery Exploitation (LERIX) - offers a new approach that combines robust GIS (geospatial Information System) methods with innovative new language-aware approaches to image and document exploitation optimized for Arabic, Farsi, and English. Our GIS capabilities represent a new approach for conflation using curvilinear features, and in particular road segments and complete road networks, for the registration of maps and imagery. Our solution uses shift-, rotation-, and scale-invariant metrics to support conflation, transforming the registration problem into a spatial similarity assessment problem. And our Language capabilities provide a novel and integrated approach to text segmentation, language recognition, text extraction, and translation.