Transfer learning-based accurate detection of shrub crown boundaries using UAS imagery

Authors: LI, JIAWEI - Orise Fellow; Zhang, Huihui; Barnard, David

Submitted to: Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/27/2025
Publication Date: 7/3/2025
Citation: Li, J., Zhang, H., Barnard, D.M. 2025. Transfer learning-based accurate detection of shrub crown boundaries using UAS imagery. Remote Sensing. 17(13). Article e2275. https://doi.org/10.3390/rs17132275.
DOI: https://doi.org/10.3390/rs17132275

Interpretive Summary: Accurately mapping the boundaries of shrub crowns is important for tracking plant life, managing land, and understanding how plants grow in fragile areas like semi-arid shrublands. Traditional image processing methods often have trouble with overlapping shrub branches, but deep learning techniques, like convolutional neural networks (CNNs), can provide better solutions for detecting them clearly. In this study, we created a method that uses deep learning models to analyze high-quality images taken by drones, including pictures in different light spectrums. This method successfully detected individual shrub crowns, even in areas with limited data. Our approach performed better than traditional methods, and showed that deep learning can work well even when data are hard to come by. This method could be helpful for monitoring the environment in regions where it's difficult to get enough data, like in remote or understudied areas. It also supports important decisions for managing fragile ecosystems and shows how deep learning can be used in situations with little data, opening up the possibility for long-term environmental monitoring and vegetation management.

Technical Abstract: The accurate delineation of shrub crown boundaries is critical for ecological monitoring, land management, and understanding vegetation dynamics in fragile ecosystems such as semi-arid shrublands. While traditional image processing techniques often struggle with overlapping canopies, deep learning methods, such as convolutional neural networks (CNNs), offer promising solutions for precise segmentation. This study employed high-resolution imagery captured by unmanned aircraft systems (UASs) throughout the shrub growing season and explored the effectiveness of transfer learning for both semantic segmentation (Attention U-Net) and instance segmentation (Mask R-CNN). It utilized pre-trained model weights from two previous studies that originally focused on tree crown delineation to improve shrub crown segmentation in non-forested areas. Results showed that transfer learning alone did not achieve satisfactory performance due to differences in object characteristics and environmental conditions. However, fine-tuning the pre-trained models by unfreezing additional layers improved segmentation accuracy by around 30%. Fine-tuned pre-trained models show limited sensitivity to shrubs in the early growing season (April to June) and improved performance when shrub crowns become more spectrally unique in late summer (July to September). These findings highlight the value of combining pre-trained models with targeted fine-tuning to enhance model adaptability in complex remote sensing environments. The proposed framework demonstrates a scalable solution for ecological monitoring in data-scarce regions, supporting informed land management decisions and advancing the use of deep learning for long-term environmental monitoring.