Project Ground Truth: Using drones and AI to automatically measure post-disaster damage

The Problem

Following a natural disaster, conducting damage assessments is a critical step to unlocking state and federal aid. However, the traditional process is slow, difficult to scale, and relies on subjective and inconsistent human classification on the ground. This operational bottleneck can significantly delay the delivery of essential aid to affected residents.

The Solution

To address this, the City of Kansas City, Missouri, partnered with X, The Moonshot Factory, on Project Ground Truth. The project, also known as Bellwether, uses a novel AI approach to act as a 'Phase Zero' first responder. Drones are deployed to capture high-resolution aerial imagery of the affected area, even while ground conditions may still be hazardous. Bellwether's AI models then automatically process these images to classify property damage according to FEMA standards, generating the necessary content to accelerate the filing of preliminary damage assessments.

The Implementation

The solution was tested in a simulated disaster exercise in Kansas City involving 60 city staff and volunteers across a 169-acre site with 120 houses showing varying levels of damage. Drones flew a pre-programmed path to capture images, which were then stitched together into a geomosaic. The Bellwether engine segmented this large image into individual parcels. An AI model then assessed each structure, assigning it a FEMA damage classification score (e.g., destroyed, major, minor, affected, intact).

The Impact

The results were a dramatic improvement over manual methods. The AI-powered process completed the damage assessment for the entire area in just 20 minutes of processing time, achieving 96% accuracy. In contrast, the equivalent manual survey method was estimated to take 78 total man-hours. The project successfully replaced subjective ground surveys with deterministic machine classification, accelerating the assessment process from days to minutes.

The project was designed to automate and accelerate the filing of FEMA (Federal Emergency Management Agency) preliminary damage assessments, implying adherence to its classification standards.