Recent technological advancements in uncrewed aerial vehicles (“UAVs” or “drones”), computer vision, and image processing have opened a new era in which forests can be mapped at the individual tree level across broad extents using imagery from low-cost consumer drones. The Open Forest Observatory (OFO), based at the University of California Davis, has developed a set of open-source software tools that make it easy for forest scientists and practitioners to employ advanced image and geospatial data processing algorithms to detect individual trees in drone imagery, map them geospatially, classify their health status and species using computer vision, and evaluate the accuracy of the predictions against co-located ground-based inventory data. This workshop will guide participants through a forest mapping and accuracy assessment workflow via a series of hands-on computer activities employing OFO tools and external open-source software packages. Participants will have the option to use each tool via command-line commands or in Python via Jupyter notebooks.

The workshop will begin with a drone-derived forest imagery dataset already processed using structure-from-motion photogrammetry into key products: an orthomosaic, a canopy height model, a 3D mesh model, the estimated camera lens calibration, and the estimated pose of the camera for each photo. The workshop will apply the following OFO tools in sequence, with each segment including a conceptual discussion followed by a hands-on walkthrough.

Tree Detection Framework (TDF): This Python package provides a standardized user interface to a range of existing models and algorithms for individual tree detection from drone-derived photogrammetry products. When applying tree detection to geospatial data, large orthomosaics must be split into small "chips" for input to the computer vision model, and the resulting chip-level predictions must be reassembled. TDF provides this "geospatial boilerplate" functionality via a standardized interface in which the user specifies chip resolution, dimensions, and stride the same way regardless of which tree detector they use. It also implements a popular geometric algorithm for detecting treetops as local maxima in a canopy height model, via the same interface, enabling easy and rigorous intercomparison. Participants will use TDF to run multiple approaches for detecting and delineating individual trees from the drone-derived orthomosaic and/or canopy height model.

Geograypher: When classifying tree species from drone imagery using computer vision models, the most common approach is to use the orthomosaic, which provides a single top-down view of each tree. This approach ignores the wealth of information in the raw drone images, which are highly overlapping and therefore provide numerous distinct views of each tree from different angles. However, drone images are not geospatial data products, and there is no direct way to translate the locations of tree crowns in raw drone images into precise geospatial polygons. Geograypher fills this gap. It leverages the fact that when you know the precise position and orientation of the drone camera, along with the camera's lens model, each pixel in a raw drone image can be translated into a 3D geospatial ray. Combined with the 3D mesh model derived from photogrammetry, these rays can be translated into precise geospatial points. Geograypher employs these concepts, drawing on computer graphics principles, to render geospatial information onto raw drone images and project information from those images into geospatial coordinates. Participants will use Geograypher to render geospatial drone-detected tree crowns onto raw drone images and produce a cropped chip for each view of each tree to supply to a computer vision model.

Tree Classification Framework: This tool provides a simple interface for loading a pre-trained computer vision classification model and running it to predict the class (e.g., species or health status) of trees in images cropped to the tree crown. Participants will use this tool to access pre-trained OFO species classification and health status (live/dead) computer vision models and apply them to the cropped tree images produced in the previous step in order to obtain predicted classes for each geospatial tree crown polygon.

Tree Registration and Matching (TRAM): The gold standard for drone-based tree mapping accuracy assessment is comparison against manual ground-based geospatial forest inventory data. However, due to systematic (site-level) and random (tree-level) geospatial mapping errors, drone and ground datasets never precisely align. TRAM identifies the optimal x-y shift to apply to a ground-based tree map to align it with a drone-derived map. It also applies heuristics to determine which trees from the drone-derived map match which trees from the ground-based map, enabling computation of accuracy statistics such as precision, recall, and F-score. Participants will use TRAM to evaluate their drone-derived tree maps against co-located OFO ground reference data.