A Unified Framework for Plant Identification from UAV Images

Unmanned Aerial Vehicles (UAVs) with their ability to capture imagery at various spatial and temporal scales offer significant potential for applications in plant identification and biodiversity monitoring. In this paper, we propose a unified framework for plant identification from UAV images comprising three components: orthomosaic image construction, plant segmentation, and plant identification. First, orthomosaic images are generated from images captured by UAVs using OpenDroneMap. Then, these images are segmented into forest, deforestation, and other areas. Finally, image patches that are extracted from forest regions are fed into plant identification to determine the plant species. Extensive experiments were conducted to evaluate the effectiveness of each component on a self-collected dataset captured from Ben En National Park in Thanhhoa province. Preliminary results demonstrate the potential application of UAVs in biodiversity mapping.