2026-09-03 –, Room3
—An Observation-Aware, Offline-First System for Repeatable Field-to-Tree Production—
This talk presents MIMAR, an observation-aware, offline-first production architecture built on OpenDroneMap and OpenSfM. It connects controlled UAV acquisition, high-recall reconstruction in repetitive canopies, GPU-accelerated matching and bundle adjustment, recoverable local execution, spatial quality gates, and decision-ready tree records.
The focus is not a faster component, but a repeatable field-to-tree system whose performance is accepted only when matching recall, reproducibility, mission geometry, and operational economics survive together.
MIMAR is an independent, offline-first architecture built on OpenDroneMap, OpenSfM, OpenMVS, open-source compute, and NVIDIA GPU acceleration for difficult UAV missions in tropical plantations.
Repetitive canopies, uncertain pose, and changing field conditions create ambiguous matches and fragile tracks. MIMAR treats solar time, illumination, weather, GSD, overlap, and sidelap as an observation contract. Position and pose are controlled priors; image evidence, robust geometry, and bundle adjustment determine whether the mission remains connected. High-recall GPU matching preserves evidence; accelerated sparse BA supports mission-scale geometry.
Raw imagery is processed locally in version-locked Docker environments with QA gates, evidence logs, and resumable recovery states. Validated geospatial and tree-level products move only when connectivity is available.
MIMAR is not a new SfM algorithm. Its contribution is a controlled field-to-tree production architecture: components are accepted only when recall, reproducibility, mission geometry, recovery, and decision value survive together under field, compute, and business constraints.
OpenDroneMap documentation: https://docs.opendronemap.org/
OpenSfM documentation: https://opensfm.org/docs/
OpenMVS project and documentation: https://github.com/cdcseacave/openMVS
FAISS documentation: https://faiss.ai/
PDAL documentation: https://pdal.io/
GDAL Cloud Optimized GeoTIFF documentation: https://gdal.org/en/stable/drivers/raster/cog.html
OpenDroneMap (ODM), OpenSfM, OpenMVS, PopSift, FAISS, PDAL, GDAL, and Docker
I make my conference contribution available under the CC BY 4.0 license. The conference contribution comprises the abstract, the text contribution for the conference proceedings, the presentation materials as well as the video recording and live transmission of the presentation:
A geospatial and remote sensing engineer, specializing in scalable and viable systems that integrate product design, production engineering, and operational efficiency.