Plot Finder: An algorithm for automatically detecting plots in UAV images

Abstract

Field-level phenotyping is a foundational component of plant breeding but remains a time-and labor-intensive task. Recent advances in imaging sensors and unmanned aerial vehicles (UAVs) have enabled high-throughput data collection at the field scale. However, phenotypic analysis is typically conducted at the plot level, requiring individual plots to be accurately extracted from UAV imagery. Because field experiments often contain hundreds to thousands of plots, this extraction step has become a major bottleneck in UAV-based phenotyping workflows. In this work, we present an automated algorithm for plot extraction from UAV images. We evaluate the method across eight datasets spanning three crop types (corn, potato, and beet) and demonstrate that our approach outperforms existing methods.