AI and 3D geospatial platform for forest resource management

A licensed forest production enterprise authorised by the Vietnamese Department of Agriculture and Forestry was relying on ground survey teams to count trees, assess health, and plan harvests – time-consuming, costly, and often risky work in remote terrain.

The enterprise also had to comply with strict forestry regulations: tree-counting principles tied to carbon credits, mandatory regeneration and protection requirements, and standardised reporting on tree density and forest health.

We built a regulation-aligned AI 3D forestry platform with three integrated layers: a 3D Forest Picture Layer that creates an accurate spatial representation of the forest from aerial imagery; an AI Analytics Layer for tree-level intelligence and forest-level indicators; and a Testing Procedures layer that covers compliance, transparency, and decision support.

The pipeline runs: Field Data Acquisition → Spatial Processing & 3D Modeling (DSM/DTM/Ortho) → AI Segmentation Engine → Data Validation & Management → Analytics & Visualisation, with an accuracy feedback loop for model retuning.

Tech stack: PyTorch, TensorFlow, MLflow, Databricks, Apache Airflow, AWS.

70% reduction in manual effort and 92% accuracy on tree inventory, with 35.1% lower labelling costs versus the client’s previous AI solution.

The reusability of the data and AI engine improved by ~60% across different tree types, giving the enterprise a scalable foundation that extends beyond the initial deployment into new forest lots and carbon-credit reporting.