Detection and Monitoring of Woody Vegetation Landscape Features Using Periodic Aerial Photography

Woody vegetation landscape features, such as hedges, tree patches, and riparian vegetation, are important elements of biotic diversity. For the reason that biodiversity loss is one major ecological problems in EU, it necessary to establish efficient workflows for registration monitoring woody features. In paper, we propose evaluate a methodology automated detection changes features from digital orthophoto (DOP). We demonstrate its ability capture most actual field thereby provide valuable support more maintenance feature layers, which shaping future environmental policies. While reliable source cover mapping combination LiDAR high-resolution imagery, can be prohibitively expensive continuous updates. The DOP cyclic aerial photography presents an alternative up-to-date information tracking in-between recordings. proposed uses segmentation neural network, trained with latest against last known ground truth target. output layer detected changes, validated by user before being used update layer. was tested using data typical traditional Central European cultural landscape, Goričko, north-eastern Slovenia. achieved F1 per-pixel 83.5% 77.1% two- five-year differences between LiDAR-based reference DOP, respectively. validation at minimum area threshold 100 m2 percentage 20% showed model recall close 90%.