Monitoring the extent of plateau forests has drawn much attention from governments given the fact that the plateau forests play a key role in global carbon circulation. Despite the recent advances in the remote-sensing applications of satellite imagery over large regions, accurate mapping of plateau forest remains challenging due to limited ground truth information and high uncertainties in their spatial distribution. In this paper, we aim to generate a better segmentation map for plateau forests using high-resolution satellite imagery with limited ground-truth data. We present the first 2 m spatial resolution large-scale plateau forest dataset of Sanjiangyuan National Nature Reserve, including 38,708 plateau forest imagery samples and 1187 handmade accurate plateau forest ground truth masks. We then propose an few-shot learning method for mapping plateau forests. The proposed method is conducted in two stages, including unsupervised feature extraction by leveraging domain knowledge, and model fine-tuning using limited ground truth data. The proposed few-shot learning method reached an F1-score of 84.23%, and outperformed the state-of-the-art object segmentation methods. The result proves the proposed few-shot learning model could help large-scale plateau forest monitoring. The dataset proposed in this paper will soon be available online for the public.

As a crucial high-altitude habitat and a key stopover site for migratory birds, Bird Island on Qinghai Lake requires automated monitoring for ecological conservation. In this context, bird detection poses a significant challenge due to targets’ large scale variations, morphological similarities, and complex backgrounds. Current general-purpose detection models struggle to adequately perceive such fine-grained features, leading to high rates of missed and false detections in complex natural scenes. Based on the YOLOv8 architecture, this study introduces two core improvements: a High-Frequency and Spatial Dependency Perception module to enhance multi-scale feature extraction, and an adaptive knowledge-base system that incorporating ornithological knowledge to boost discrimination among similar species via knowledge-guided inference. The resulting model is named YOLO_BD (You Only Look Once_Bird Detection). Experimental results demonstrate that YOLO_BD achieves a mean Average Precision (mAP@0.5) of 75.2% with 6.6 million parameters while maintaining high detection efficiency. It not only surpasses the baseline YOLOv8n (72.8% mAP@0.5 with 3.1 M parameters) in accuracy but also outperforms the larger YOLOv8s model (74.4% mAP@0.5 with 11.1 M parameters), highlighting its dual advantages in achieving lightweight design and enhanced performance. This research provides an effective technical solution for intelligent wildlife monitoring in resource-constrained environments.