OUSD (R&E) CRITICAL TECHNOLOGY AREA(S): Trusted AI and Autonomy OBJECTIVE: Develop a physical damage detector for overhead imagery to prioritize the collection of additional imagery and the allocation of damage assessment analysis resources. DESCRIPTION: **NGA has designated the automated damage assessment and change detection algorithm developed for this topic as Controlled Unclassified Information (CUI). The technical solution shall include a plan for protecting the weights and technical specifications associated with the automated damage assessment or change detection algorithm from unauthorized access and theft. DFARS clause 252.204-7012 applies.** This announcement seeks technologies that automate physical damage assessments using commercial remote-sensing data. Assessing physical damage of infrastructure, buildings, etc., over large geographic areas due to human conflict or natural disasters requires a substantial amount of human labor. Timely damage assessments are necessary for the effective allocation of resources, such as for humanitarian assistance and disaster recovery (HADR), and providing situational awareness to military commanders and policy makers. Overhead commercial satellites and airborne sensors are frequently used by geospatial and imagery analysts to assess physical damage [1, 2]; however, few commercial sensors have associated automated damage assessment capabilities fine-tuned to that sensor with clearly defined performance bounds across geographic areas, types of damage, and types of structures. The aim of this effort is to design and develop capabilities that reliably identify physical damage to buildings and infrastructure that meet the following criteria: 1. 1 .Solutions apply to at least one commercial overhead sensor, where proposers shall identify their own source of remotely sensed data, i.e., the government will not necessarily furnish the data. 2. Output of the solution, e.g., polygons and associated metadata, can be ingested into geographic information system (GIS) software platforms using commonly used data formats, such as KML [3]. 3. Damage estimates are accompanied by meaningful and human-interpretable confidence values, e.g., like the expressions of likelihood detailed in Intelligence Community Directive (ICD) 203 [4]. 4. If more than two levels of physical damage are inferred, as opposed to the binary classes of undamaged vs. damaged/destroyed or change detection, then the damage levels align with common production standards (e.g., CJCSI 3162.02 for Department of Defense customers [5]). 5. Solution and sensor operating bounds (e.g., geographic area limitations of the solution or sensor, types of damage or effect, collection geometries, time of day) clearly defined, along with confidence estimates of those bounds, where qualitative estimates could be acceptable with proper justification. 6. Evaluation measures of performance demonstrated on publicly available data. If only proprietary data are available for evaluation, then the evaluation data must be thoroughly described to make the comparison of performance across commercial vendors possible. In either case, the data must be sufficiently described to understand the scope of the evaluation and whether there is sufficient evidence that a solution would be expected to generalize to other areas or damage types. 7. If a two-stage approach of building segmentation, or more generally feature segmentation, followed by damage inference is used or proposed, then proposers are strongly encouraged to use commercially or openly available feature polygons. A strong justification will be needed if not using a commercially or openly available feature polygon data set, e.g., no data sets exist for a given test area. If a feature polygon data set of buildings, non-building structures, etc., is created as part of this effort, then the National System for Geospatial Intelligence (NSG) Application Schema definitions (NAS) [6] and GEOINT Content Extraction Specification (GCES) [7] should be used, at least as a baseline. Deviation from the NAS and GCES is acceptable if the proposer articulates where the differences are with sufficient justification. A reasonable assumption for usage of these capabilities is that the output will be used to identify potential physical damage for further human review, where the goal is to reduce the amount of human labor required to perform damage assessments. The assumption that outputs will be used in human-machine teams can be used to determine how the criteria in the topic description are addressed, including how to communicate operating bounds, target evaluation measures, etc. PHASE I: Research and design of an automated damage assessment or change detection solution related to physical damage following the criteria in the topic description. Proposals must demonstrate an innovative approach, whether that be enabling an underutilized sensor or phenomenology and therefore demonstrating a strong understanding of the trade space, proposing a novel solution that advances the state of the art, or measurably and substantially reducing human labor required for damage assessments over wide geographic areas. Deliver a written report on the design, any preliminary results, and a Phase II plan. This topic is accepting both Phase I and Direct to Phase II (DP2) proposals. Proposers interested in submitting a DP2 proposal must provide documentation to substantiate that the scientific and technical merit and feasibility described above has been met and describe the potential commercial applications. Documentation should include, reference, or summarize all relevant information including, but not limited to technical reports, test data, prototype designs/models, and performance goals/results. PHASE II: Implement an automated damage assessment or change detection algorithm related to physical damage following the criteria in the topic description and Phase I expectations. Effective implementations for all criteria are expected. Deliver a written report detailing results. PHASE III DUAL USE APPLICATIONS: If proposed solutions are developed with or derived from Open-Source code or data, the Open-Source licenses of those original sources must allow for commercialization and the proposed use documented in the SBIR proposal. Data used to train an algorithm should include proper citations and sourcing, including licensing and releasability. REFERENCES: 1. D. Palumbo, A. Abutaleb, P. Cusiac, and E. Rivault, At least half of Gaza's buildings damaged or destroyed, new analysis shows, BBC, January 30, 2024. [Online]. Available: https://www.bbc.com/news/world-middle-east-68006607. [Accessed Jan. 16, 2025]. 2. Y. Asi, D. Mills, P. G. Greenough, D. Kunichoff, S. Khan, J. Van Den Hoek, C. Scher, S. Halabi, S. Abdulrahim, N. Bahour, A. K. Ahmed, B. Wispelwey, and W. Hammoudeh, 'Nowhere and no one is safe': spatial analysis of damage to critical civilian infrastructure in the Gaza Strip during the first phase of the Israeli military campaign, 7 October to 22 November 2023, Conflict and Health, vol. 18, no. 24, 2024, https://doi.org/10.1186/s13031-024-00580-x. 3. KML, Open Geospatial Consortium. [Online]. Available: https://www.ogc.org/publications/standard/kml/. [Accessed Jan. 27, 2025]. 4. Analytic Standards, Intelligence Community Directive 203 Technical Amendment, Office of the Director of National Intelligence, 2023. [Online]. Available: https://www.dni.gov/files/documents/ICD/ICD-203.pdf. [Accessed Jan. 16, 2025]. 5. Methodology for Combat Assessment, United States Joint Chiefs of Staff, 2021. [Online]. Available: https://www.jcs.mil/Portals/36/Documents/Library/Instructions/CJCSI%203162.02A.pdf. [Accessed Jan. 16, 2025]. 6. Building, NSG Standards Registry Identifier 100083, National Geospatial-Intelligence Agency, Mar. 2007. [Online]. Available: https://nsgreg.nga.mil/as/view?i=100083 7. Building, Foundation GEOINT Standards Data Product Specification Identifier 100083, National Geo-Spatial Intelligence Agency. [Online]. Available: https://fgs-dps.gs.mil/content/100083 KEYWORDS: physical damage assessment; PDA; battle damage assessment; BDA; volumetric change; change detection
