Attch1_FDT_SOO_Auto_Feature_Extraction_20260130.pdf

UNCLASSIFIED

National Geospatial-Intelligence Agency (NGA)

Foundation Digital Twin (FDT)

Automated Feature Extraction (AFE)

DRAFT Statement of Objectives (SOO)

Version 1.0

30 January 2026

CHANGE HISTORY

Date Section

Changed

Reason for Change Author Revision

2/19/26 Initial Issue/RFI Release J. Martin 1.0

1.0 Summary and Background

The overall purpose of this Statement of Objectives is to acquire a contractor to provide a complete, end-to-end Automated Feature Extraction (AFE) solution by identifying common features (Appendix A) and detailing individual use case for the feature(s) (Section 3.0). This solution will modernize NGA's geospatial intelligence production by leveraging AI/ML to automate manual processes, improve data accuracy, and ensure seamless integration with critical NGA production systems. The successful implementation of this AFE capability will significantly enhance the timeliness, relevance, and accuracy of geospatial intelligence, directly supporting national security missions.

The National Geospatial-Intelligence Agency (NGA) is a Department of War (DoW) combat support agency and a member of the National Intelligence Community (IC). NGA develops imagery and map-based intelligence solutions for U.S. national defense, homeland security, and safety of navigation.

NGA’s mission is to “provide timely, relevant, and accurate geospatial intelligence in support of national security.”

The current Foundation GEOINT environment faces significant operational challenges that impede efficiency and effectiveness. Foundation GEOINT (FG) Analysts spend extensive time on repetitive manual feature extraction, conflation, and validation, which are time-consuming activities that could be automated.

To address these challenges and support NGA's 2035 GEOINT Concept of Operations, NGA is implementing the AFE capability. This initiative will leverage artificial intelligence and machine learning

(AI/ML) to transform geospatial intelligence production, significantly reducing manual processing time, increasing data accuracy, and enabling the rapid delivery of intelligence to decision-makers.

2.0 Scope of Work

The scope of this effort is to acquire services or capabilities for the full lifecycle of the AFE capability.

This includes the design, development, integration, deployment, and sustainment of the AFE capability to automate the identification and extraction of geospatial features from imagery and other source materials.

The AFE capability will be a critical enabler for modernizing Foundation GEOINT processes and will be integrated with key NGA systems. The work will support NGA's maritime, aeronautical, topographic, geographic, and geomatics mission domains and will be integrated into NGA’s FG production environments on all required security domains.

3.0 Use Cases

This section details key use cases that exemplify NGA’s operational requirements for an Automated

Feature Extraction (AFE) solution. Responding vendors shall select one, two, or more features from the comprehensive list provided in Appendix A. For each feature selected, the vendor must describe their solution's capabilities in the context of the most relevant use case category outlined below. The response should specifically detail the solution's proficiency across the three primary AFE functions: (A) Object

Detection, (B) Geometry Extraction, and (C) Attribution, as defined in this section.

AFE RFI Features and Categories individual use cases:

1) Buildings

a. Including agglomerations as Facility, Military Installation, or Built-Up Area

b. Rationale (internal): The extraction of building geometries is well established with multiple different approaches. However, NGA has not settled on a specific approach/vendor. And most approaches just get geometries but collect only limited attribution/characterization information about those buildings. Because of this it’s useful to keep buildings by themselves as a category.

c. Example Use Case: Identify all the built structures in a city and identify those that are places of worship.

d. Use Case Objectives:

i. Global building footprints that accurately represent individual buildings at their ground-level footprint in vector format (e.g., splitting polygons along shared walls, polygons match the exterior walls).

ii. Characterization of Buildings into functional types aligning with TDS v7.1 and

MGCP TRD functional categories.

iii. Methods to associate buildings with each other into facilities, installations, or built-up areas agglomerations.

2) Linear Transportation Features (Road, Railway, Cart Track, Trail)

a. Rationale (internal): Long linear features present unique, but associated challenges for

CV algorithms. The ontological differences between a Road, Cart Track, and Trail are often unclear so approaching this problem as a group is likely more efficient than trying to overly parse by feature type.

b. Example Use Case: Determine potential transportation routes between Point A and Point

B.

c. Use Case Objectives:

i. Extract a consistent, topologically correct transportation network that reflects the real-word transportation system as linear vector geometry, either along road centerlines or along ways of travel.

ii. Characterize physical attributes of the transportation features that influence navigability (e.g., width, surface composition, weather restriction).

iii. Characterize attributes about Transportation Feature use dynamics that are observable from behavior patterns (e.g., roadway type, one-way streets, lane counts, speed limits).

3) Discrete Transportation Features (Bridge, Ford, Tunnel, Gate)

a. Rationale (internal): Transportation networks are typically most restricted by discreate features that limit what vehicles are capable of navigating the network. These feature all need to be associated with the network but are notionally independent of the transportation features and can use different detection methods.

b. Example Use Case: Determine if a vehicle (with specific height and weight) can navigate through choak points between Point A and Point B.

c. Use Case Objectives:

i. Identify and extract vector geometries (points, lines, and polygons) that represent real-world features that are associated with linear transportation ways.

ii. Characterize attributes of these features that influence successful transportation through these features (e.g., clearance height, load capacity, width)

iii. Associate these features with separately generated linear transportation geometries (e.g., have the road go through the tunnel or on the bridge).

4) Utilities (Cistern, Pylons, Cables, Electric Power Station, Pipeline, Power substation, Water

Tower, Water Well)

a. Rationale (internal): Utilities consist of both discrete and connected linear features but are all in association. In traditional manual collection many of these features can only be identified because of these associations (e.g., cables and pylons). For this reason, we can look at these as a collective problem-set.

b. Example Use Case: Determine what utilities connect to and serve a particular facility.

c. Use Case Objectives:

i. Identify and extract connected topologically correct utility networks (power, water, communications), with particular focus on longer distance services (e.g., high-voltage electrical cables rather than residential service lines).

ii. Characterize the attributes of extracted features, especially their association with certain types of utilities.

iii. Characterize the physical attributes (esp. height) of these features.

5) Built Water Infrastructure (Ditch, Culvert, Inland Waterbodies, Dam)

a. Rationale (internal): Natural water features are collected through traditional LULC image classification efforts. But many of the constructed water features do not get picked up or get classified as simply “Water.” Because hydrological features are all affected by elevation they can often be identified using digital elevation models in conjunction with imagery. They are also frequently identifiable through associations with each other (e.g., a dam and a reservoir).

b. Example Use Case: Identify all dams in a region.

c. Use Case Objectives:

i. Identify and extract vector representation (point, line, polygon) of these built water features.

ii. Where associated with existing features, put into correct spatial relationship (e.g., culverts and road, dams and waterbodies).

iii. Characterize the physical attributes of these features.

6) Aeronautical Infrastructure (Helipad, Runway, anything else SFA might need)

a. Rationale (internal): SFG generally collects minimal aeronautical features but does need to maintain some for display on topographic map products. These are typically disconnected asphalt or concrete areas that traditional LULC classifications identify simply as built areas, without further classification.

b. Example Use Case: Identify runways in association with the ground transportation network for mixed air-land transportation planning.

c. Use Case Objectives:

i. Identify and extract vector representations of on-ground aeronautical features.

ii. Characterize the physical attributes of these features.

THREE PRIMARY AFE FUNCTIONS

A) Object Detection (finding the object)

a. This is the standard computer vision capability for identifying that something exists in an image, either providing a return as a point, bounding box, or other form that does not represent the real-world dimensions of the object.

B) Geometry Extraction (creating a vector representation of the object)

a. This is typically what we mean by AFE and is a follow-on from A. This is the correct extraction of the feature from the imagery representing its real-world dimension in accordance with some extraction standards (e.g., the centerline of a road, the footprint of building, the roofline of a building).

C) Attribution (characterizing and providing information about the object)

a. This is the addition of information to a geometry that reflects informative characteristics about the feature. It can overlap with A based on how fine-detailed the feature type is that is being detected (e.g., a detector for Buildings would need an attribute to determine if it is a hospital, but a detector for Hospitals would implicitly have that attribute). Typically, which attributes are needed is driven by the data schema used for each mission set.

i. This could be further sub-divided into categorization (e.g., hospital, factory) and measurement (e.g., height, width).

4.0 Objectives

The contractor shall achieve the following five high-level objectives to successfully deliver a comprehensive Automated Feature Extraction (AFE) solution:

Objective 1

Develop and Deploy a High-Performing Automated Feature Extraction (AFE) Capability

The primary goal is to develop, deliver, and sustain a robust AFE capability that automatically detects and extracts geospatial features from various sources including imagery (EO, SAR, HS, MS) and raster maps.

This system must be capable of both monoscopic and stereoscopic extraction and handle individual and batch processing. The capability shall automatically attribute extracted features based on observable characteristics in accordance with NGA domain standards and preserve feature geometry, spatial relationships, and metadata from the source data.

Objective 2

Ensure High-Quality Data and Enable Analyst-in-the-Loop Validation

The contractor shall implement a system that ensures the accuracy and reliability of all extracted data.

This includes achieving a threshold accuracy of 90% for topographic and maritime features and 90% for aeronautical features, with an objective accuracy of 99%. The system must feature automated quality assessment processes with confidence scores and provide an intuitive interface for analysts to review, validate, and correct extracted features. All analyst feedback shall be used to continuously improve the performance of the AI/ML models.

Objective 3

Implement Automated Change Detection to Maintain Data Currency

The contractor shall provide a solution that automates the detection of changes by comparing new source data against existing feature databases. The system must be able to identify modifications, new features, and other discrepancies, generating prioritized alerts for analyst review. This will ensure that the

Foundation GEOINT data remains current and accurate, enabling timely updates and informed decision-making.

Objective 4

Achieve Seamless Integration with NGA's GEOINT Ecosystem

The AFE capability must be fully integrated into the Foundation GEOINT Engine (FGE). This includes seamless integration with the Vector & Cartographic Production System (VCPS) and the Foundation

Digital Twin (FDT) data environment for feature registration and storage. The contractor shall implement open APIs and ensure that all extracted features are stored in standardized formats (e.g., GDB, Shapefile) to support interoperability with other NGA systems and direct dissemination to customer platforms.

Objective 5

Meet and Maintain Stringent Performance and Security Standards

The solution must be scalable, reliable, and secure. The system shall be capable of processing a standard

1GB source file in less than 10 minutes and achieve 99.9% availability in the production environment.

The contractor must ensure the system is deployable to both Unclassified and Classified domains and complies with all applicable security directives, including DNI ICD 503 and all requirements outlined in the DD Form 254.

Appendix A: Prioritized Geography Features for AFE Automated Feature Extraction Features of Interest:

Automated feature extraction consisting of features represented by points, lines, and polygons. Collection of topographic features is global. Full specifications describing these features and their associated attributes are available for each data program and should be consulted in the development of AFE processes.

Generally, features are collected to the most complex geometry visible at the scale of collection.

Prioritized Features consist of the following list below. This list is an amalgam of TDS and MUVD collectable features and provides a representative sample of the feature types collected in these topographic data programs. Note that some feature types may lump together significant and important categories of interest that are typically distinguished by attributes (e.g., the Feature Function attribute on a

Building feature determines if a Building is a hospital or a religious building which are important categories for the topographic mission):

Prioritized features for AFE:

Bog

Bridge

Building

Built-up area

Cable

Canal

Cart track

Crop land

Culvert

Dam

Ditch

Electric power station

Elevation contour

Extraction mine

Forest

Grassland

Helipad

Inland waterbody

Island

Marsh

Military installation

Natural pool

Non-water well

Orchard

Pipeline

Power substation

Railway

Rice field

River

Road

Runway

Sand dunes

Soil surface region

Sports ground

Swamp

Thicket

Tower

Trail

Tunnel

Water tower

Water well

Appendix B: Additional Geography Features for AFE

Administrative boundary

Aeration basin

Aerial farm

Aerodrome beacon

Aeronautical radio navigation installation

Aircraft hangar

Allotment area

Amphitheatre

Amusement park

Amusement park attraction

Annotated location

Anti-aircraft artillery site

Approach lighting system

Apron

Aqueduct

Arcade

Archaeological site

Area of displaced persons

Asphalt lake

Astronomical observatory

Bandshell

Barn

Basin gate

Beach

Blast barrier

Blast-furnace

Bogie exchange site

Bollard

Botanic garden

Boundary monument

Bridge pier

Bridge span

Bridge tower

Brush

Building superstructure

Cableway

Cairn

Camp

Campsite

Cane

Caravan Park

Castle

Castle complex

Catalytic cracker

Causeway structure

Cave mouth

Cemetery

Checkpoint

Cistern

Cleared way

Conservation area

Contaminated region

Control tower

Conveyor

Cooling facility

Cooling tower

Courtyard

Crane

Crevasse

Crevice

Cultural conservation area

Cut

Cut line

Defensive revetment

Dish aerial

Display sign

Disposal site

Dragon’s teeth

Drive-in theatre

Dry dock

Embankment

Engineered earthwork

Engineered turnaround site

Esker

Excavating machine

Facility

Fairground

Farm

Fence

Ferry crossing

Ferry station

Firing range

Fish farm facility

Fish weir

Flagpole

Flare pipe

Floating barrier

Floating dry dock

Flood control structure

Ford

Forest clearing

Fortification

Fortified building

Fountain

Fuel storage facility

Gantry

Gate

Geologic fault

Geophysical data track line

Geopolitical entity

Geothermal outlet

Glacier

Golf course

Golf driving range

Grain elevator

Grain storage structure

Grandstand

Grave

Greenhouse

Harbor

Hardened aircraft shelter

Hazardous rock

Heating facility

Hedgerow

Heliport

Holding pen

Hop field

Hopper

Hulk

Hummock

Hut

Hydrocarbon products facility

Hydrocarbons field

Ice cliff

Ice peak

Ice route

Ice shelf

Icecap

Industrial farm

Industrial furnace

Installation

Lagoon

Land aerodrome

Land subject to inundation

Land transportation way

Land-water boundary

Launch pad

Light support structure

Light vessel

Lighthouse

Lock

Logging site

Lookout

Manufactured home park

Mariculture site

Market

Mass grave

Memorial monument

Mine shaft superstructure

Minefield

Mineral pile

Miscellaneous feature type

Missile site

Moat

Monument

Mooring mast

Moraine

Motor vehicle station

Mountain pass

Munition storage facility

Named location

Non-building structure

Nuclear reactor containment

Oasis

Office Park

Offshore construction

Outdoor theatre screen

Overhead obstruction

Park

Parking garage

Particle accelerator

Penstock

Pipeline crossing point

Plant nursery

Populated place

Port

Public square

Pumping station

Pylon

Qanat shaft

Racetrack

Radar station

Railway sidetrack

Railway traverser

Railway turntable

Railway yard

Rapids

Recycling site

Reef

Rig

Road interchange

Roadside rest area

Rock formation

Roundhouse

Ruins

Sabkha

Salt evaporator

Salt flat

Scoreboard

Seaplane run

Settlement

Settling pond

Sewage treatment plant

Shanty town

Sharp curve

Shear wall

Shed

Ship elevator

Shipping container

Shipyard

Shopping complex

Shoreline construction

Shoreline ramp

Sidewalk

Ski-jump

Sluice gate

Smokestack

Snow field and/or ice field

Snow ice field

Solar farm

Solar panel

Space facility

Spillway

Spot elevation

Spring

Stadium

Stair

Steep terrain face

Stopway

Storage depot

Storage tank

Structural pile

Surface bunker

Swimming pool

Tactical air navigation aid

(TACAN) beacon

Tank farm

Taxiway

Test site

Tethered balloon

Tidal water

Tomb

Training site

Transportation block

Transportation platform

Transportation route protection structure

Transportation station

Tree

Tundra

Tunnel mouth

Underground bunker

Underground dwelling

Vanishing point

Vehicle barrier

Vehicle lot

Vessel lift

VHF omnidirectional radio

(VOR) beacon

Vineyard

Void collection area

Volcano

Wall

Water aerodrome

Water intake tower

Water mill

Water race

Water treatment bed

Waterfall

Waterwork

Weigh station

Wind farm

Wind turbine

Windmill

Wreck

Zoo