Orchard Structure Classification

The Orchard Structure classification honours project for Aerobotics aimed to determine what planting patterns were present in a GeoJSON dataset of an orchard. The was a secondary goal of parameter extraction to extract information about the pattern in the dataset. The GeoJSON files outlined polygons of trees at different confidence intervals and the centroid of these polygons was used for the project. Three approaches were used: Template Matching (Jonathon Everatt), Hough Transform (Piero Puccini), and Image comparison using the Hausdorff distance (Chelsea van Coller). Hough Transform was only able to detect the rotation of the orchard in the image. Template matching could detect the pattern present and some of the key parameters. Although it struggled to differentiate between similar templates. Hausdorff distance could also detect the pattern present in the dataset but had some efficiency problems due to the issue of rotation.