Building a reliable vegetation baseline from aerial surveys is a mess of disconnected tiles and point clouds. A new open reference architecture, published under CC BY 4.0, solves this by binding 4-band imagery and LiDAR into a strict site ontology. The project, hosted at codeatoms.ai, provides a JSON object model and a model register that allows operators to reuse the structure without being locked into a specific vendor's ecosystem.

The Acquisition Is the Anchor

The architecture begins with a rigid acquisition protocol to ensure data consistency across seasons. The design mandates a single manned flight within a week either side of July 1, capturing 3-inch orthoimagery in four bands (red, green, blue, and near-infrared) alongside USGS Quality Level 1 LiDAR. The LiDAR must hit a minimum of 8 pulses per square meter, a density threshold critical for making the subsequent vegetation analysis possible.

Modeling the Mission, Not Just the Data

Instead of treating files as the primary entity, the system makes the flight mission the focal object. The object model defines sixteen typed entities, ranging from the site and its watercourses to the specific camera, LiDAR sensor, and GNSS/IMU chain. This structure ensures that every product, including the classified point cloud and QA/QC accuracy reports, traces back to the mission, the sensor, and the accuracy report behind it, creating a full lineage for the data.

Keeping Compute Inside the Operator's GIS

A key architectural decision is to avoid cloud black boxes entirely. Data enters through two adapter families into the operator's ArcGIS Enterprise, ensuring nothing runs in a cloud the operator does not control. Furthermore, the system serves no model; vegetation analysis remains with the operator's own analysts, and the model register is intentionally left empty. The survey firm prices the acquisition against the operator's task table, decoupling the data collection from the analytical compute.

Key Takeaways

  • The architecture uses a strict July 1ยฑ1 week flight window to standardize vegetation baselines.
  • Sixteen typed objects create a full lineage from sensor to final QA/QC report.
  • The design explicitly avoids cloud dependency by integrating directly with ArcGIS Enterprise.
  • No models are served; analysis stays with operator personnel to maintain control.

The Bottom Line

This is a pragmatic win for infrastructure builders who are tired of losing context between survey seasons. By treating the mission as the focal object rather than the file, it creates a durable, versioned baseline that actually survives personnel turnover.