Artificial Intelligence (AI) is reshaping how we understand our world. From disaster response and infrastructure management to environmental monitoring and national security, organizations are increasingly relying on automated systems to process vast amounts of geospatial information and turn it into actionable insight.
At the same time, the volume and complexity of that information continues to grow. Imagery, video, sensor networks, digital twins, and other data sources are generating unprecedented amounts of information that must be shared, understood, and trusted across organizations, platforms, and applications.
Meeting that challenge requires more than new technologies. It requires a common foundation that allows systems to work together, data to move seamlessly between environments, and information to retain its meaning, quality, and trustworthiness throughout its lifecycle.
That challenge was at the heart of OGC Testbed-21, a seven-month collaborative initiative sponsored by NASA and NGA. Bringing together experts from government, industry, academia, and the open-source community, the Testbed explored how open standards and open-source technologies can advance geospatial interoperability across emerging technologies.
The results of the broader initiative were showcased during the OGC Testbed-21 Demonstration on May 6–7, 2026. While the demonstrations covered a wide range of topics, three themes emerged consistently throughout the event.
1. Interoperability Now Spans Formats, Semantics, and Trust
For decades, interoperability in the geospatial world largely meant ensuring that systems could exchange data. If one application could read another application’s data, the problem was considered solved. But the demands being placed on geospatial information are changing rapidly. AI systems, digital twins, automated workflows, and increasingly complex data ecosystems require more than simple data exchange. They require systems to understand information, evaluate its quality, verify its origins, and determine whether it can be trusted.
That shift was evident throughout Testbed-21. Across the demonstrations, participants explored how interoperability can extend beyond formats and APIs to include semantics, provenance, integrity, and trust. Work on GIMI (Geospatial-Intelligence Imagery Media for Intelligence, Surveillance, and Reconnaissance), Data Quality for Integrity, Provenance, and Trust (DQ4IPT), and conformance testing all pointed toward the same conclusion: Metadata is no longer an accessory to data; it is becoming a critical part of the information itself, helping both people and machines understand where data came from, how it was created, and how it should be used.
As organizations increasingly rely on automation and AI, this broader vision of interoperability becomes essential. Systems must be able to exchange not only information, but also meaning, context, quality indicators, and evidence. Testbed-21 demonstrated that the geospatial community is beginning to build that foundation.
2. Imagery and Video Are Becoming First-Class Information Assets
Imagery has always been one of the most valuable sources of geospatial information, but it has often been treated as a file to be stored, downloaded, and viewed separately from the broader information ecosystem. Testbed-21 demonstrated a different future—one in which imagery and video become rich, discoverable, and interoperable information assets that can move seamlessly through modern workflows.
Much of this work centered on GIMI, a next-generation approach to managing imagery, video, and related geospatial media. Developed with strong participation from the defense and intelligence communities, GIMI is designed to address the limitations of legacy file-based formats and workflows that have long constrained how imagery and media are discovered, shared, and used. Instead of treating imagery as isolated files, GIMI enables media to be packaged with rich metadata, exposed through modern APIs, linked to catalogs and knowledge systems, and integrated directly into digital workflows.
Throughout the demonstration, participants showcased advances in streaming, browser-based visualization, semantic enrichment, standards-based discovery, and metadata management. Rather than treating imagery as static content, these demonstrations showed how it can be integrated into catalogs, linked to machine-readable metadata, accessed through APIs, and analyzed alongside other sources of information. The result is a model in which imagery and video become active participants in digital workflows rather than passive data products.
This evolution is particularly important as AI becomes increasingly dependent on imagery and video as primary inputs. The ability to discover, access, understand, and trust visual information at scale will be critical to future applications in intelligence, disaster response, infrastructure management, environmental monitoring, and countless other domains. By making imagery and media more discoverable, interoperable, and AI-ready, GIMI provides a foundation for the next generation of geospatial and analytic systems.
3. Trust, Quality, and Conformance Are Prerequisites for AI and Automation
Perhaps the strongest message to emerge from Testbed-21 was that trust in data and analytics can no longer be assumed. As organizations increasingly rely on automated systems and AI-assisted decision-making, understanding where data originated, how it was processed, and whether it conforms to accepted standards becomes increasingly important. Without those assurances, even the most sophisticated analytics can quickly lose credibility.
The demonstrations highlighted several complementary approaches to addressing this challenge. Participants explored machine-readable provenance, integrity services, verifiable credentials, quality metadata, and automated conformance testing. Together, these capabilities help create an environment in which trust is built into the system rather than added afterward through manual review and validation. They also make it possible for organizations to evaluate data and services more consistently across complex, distributed environments.
The work reinforced a simple but increasingly important principle: trustworthy AI depends on trustworthy data, trustworthy metadata, and trustworthy implementations. As geospatial systems become more automated and interconnected, quality, provenance, integrity, and conformance will become foundational infrastructure rather than optional enhancements.
What Was Demonstrated?
Advancing GIMI: A Modern Foundation for Geospatial Imagery
A major focus of Testbed-21 was the advancement of GIMI, an emerging approach for managing imagery and video in modern geospatial workflows.
The work is already producing tangible results. Just days after the Testbed-21 Demonstration, ISO published HEIF Edition 3 Amendment 2, formally incorporating the “tili” feature into the HEIF standard. The feature, which enables more efficient access to large tiled imagery, was proposed in the prior Testbed-20 and advanced through work conducted in Testbed-21. The incorporation of the “tili” feature into the HEIF standard represents a concrete example of how collaborative research can influence international standards.
Joan Maso Pau (Universitat Autònoma de Barcelona – CREAF) presented the baseline GIMI specification established for the Testbed, defining modular requirements for file structure, content identifiers, atomic time encoding, imagery classes, security markings, and RDF-based metadata. He also highlighted related work supporting GeoHEIF, an emerging candidate OGC Standard.
The Testbed’s work on Seek-Optimized HEIF has since achieved an important milestone with the publication of ISO HEIF Edition 3 Amendment 2, which formally incorporates the “tili” feature into the HEIF specification.
Building an Open-Source GIMI Ecosystem
Several participants demonstrated how open-source software is rapidly evolving to support GIMI.
Dirk Farin (Algorithmic Research e.K.) showcased enhancements to the libheif library, including support for inter-frame video coding, improved lossless and uncompressed codecs, GeoTIFF-to-GIMI conversion, RDF embedding, and HTTP-range-based streaming of tiled imagery.
Daniel Morin (Collabora) demonstrated enhancements to the GStreamer framework that enable complex geospatial sensor data—including multispectral imagery, floating-point values, and scientific data formats—to be processed, encoded, and streamed using GIMI workflows.
Núria Julià Selvas and Joan Maso Pau (Universitat Autònoma de Barcelona – CREAF) demonstrated enhancements to the MiraMon Map Browser, enabling fluid web-based visualization of very large GIMI and Seek-Optimized HEIF datasets through HTTP range requests while maintaining access to rich metadata and semantic information.
Making Imagery Discoverable
Discovery and accessibility were recurring themes throughout the demonstrations.
Shruti Suresh and Alex Bostic (Voyager Search) demonstrated how Voyager Search has been extended to support OGC API Records and STAC, enabling GIMI imagery and RDF metadata to be cataloged, discovered, and accessed through standards-based interfaces. The demonstration highlighted how provenance information and metadata can remain connected to imagery assets without requiring proprietary tooling.
Bringing Semantics to Geospatial Media
Simon Cox presented work on an ontology for coverage description that explored how GIMI metadata can be formally represented using OWL while aligning with established semantic frameworks including Basic Formal Ontology (BFO), Common Core Ontologies (CCO), GeoSPARQL, and Semantic Sensor Network (SSN) ontology.
This work lays an important foundation for validation, machine understanding, and automated reasoning across geospatial systems.
Connecting Video, Location, and Analytics
Rob Smith (Away Team Software) demonstrated Web Video Map Tracks (WebVMT) using dash-cam footage synchronized with high-frequency location and accelerometer data.
The prototype showed how browser-based applications could identify collision-like events, potholes, and other anomalies directly from georeferenced video streams, illustrating the growing role of video as a geospatial information source.
Evaluating Performance at Scale
Eugene Yu (George Mason University) presented preliminary performance evaluation results indicating that GIMI can achieve performance comparable to Cloud Optimized GeoTIFF when properly encoded.
Martin Desruisseaux (Geomatys) followed with a demonstration of work underway within Apache SIS to benchmark GIMI against other imagery formats and support future implementation guidance.
Núria Julià Selvas (Universitat Autònoma de Barcelona – CREAF) concluded the GIMI session with a summary of progress achieved during the Testbed, including proposals for Seek-Optimized HEIF and harmonization efforts involving GIMI, GeoHEIF, and datacube technologies.
Establishing Trust Through Integrity, Provenance, and Quality
The second day focused on Data Quality for Integrity, Provenance, and Trust (DQ4IPT).
Lucio Colaiacomo (4113 Engineering) presented approaches for associating geospatial data products with machine-readable evidence describing lineage, processing history, quality characteristics, and intended use.
Andreas Matheus (Secure Dimensions) and Yves Coene (Spacebel) demonstrated a suite of IPT services that enable evidence artifacts to be published, discovered, validated, and exchanged using open standards. The demonstration combined OGC APIs, STAC, verifiable credentials, and blockchain-based technologies to illustrate how trust services can operate across an interoperable ecosystem.
Jamie Feiss (KurrawongAI) demonstrated a software library that enables transformation of data-quality metadata between standardized models, helping organizations integrate quality information into operational workflows.
Conformance Testing and Developer Enablement
The final demonstrations focused on implementation quality and interoperability assurance.
Charles Heazel (Heazel Technologies) presented the GIMI Conformance Testing Tool developed through the Testbed, while Samantha Lavender (Pixalytics) presented accompanying training materials designed to help developers adopt and use the tool.
Gobe Hobona (OGC) and Samantha Lavender then demonstrated how automated conformance testing can assess implementations against GIMI requirements, helping developers identify issues early and increasing confidence that independent implementations will interoperate as intended.
Looking Ahead
The Testbed-21 Demonstration showcased far more than a collection of prototypes. It provided a glimpse into how the geospatial ecosystem is evolving to support a future defined by AI, automation, digital twins, and increasingly complex information environments. Across both days, participants demonstrated practical advances in imagery formats, semantics, provenance, quality assessment, trust frameworks, and developer tooling—advances that are helping build the foundations of a more connected and intelligent geospatial ecosystem.
Just as importantly, the event highlighted the unique value of OGC’s collaborative Testbed model. By bringing together organizations with different perspectives and expertise, Testbeds create a space where ideas can be explored, technologies can be validated, and implementation challenges can be addressed before they reach operational deployment. The work showcased during Testbed-21 demonstrated not only what is possible, but what becomes achievable when government, industry, academia, and the open-source community collaborate around shared challenges. As these efforts mature, they will help ensure that the next generation of geospatial systems is not only interoperable, but also trustworthy, discoverable, and ready to support increasingly sophisticated forms of analysis and decision-making.
Recordings
The recording from Part 1 of Day 1 is available here: https://files.ogc.org/external/de89426fe5f6d20d7546d153b20fb12aadbb02e6ffd6c7cb3ea1e5b4644783de
The recording from Part 2 of Day 1 is available here: https://files.ogc.org/external/file/f5xm957cad2a272af4241abd5901cab04bcd6
The recording of Day 2 is available here: https://files.ogc.org/external/file/w69tpd4866fb9f0be4e1e937674dcd2977de0
