Pantholops hodgsonii functions as a keystone taxon within plateau ecosystems, yet its harsh high-altitude habitat imposes substantial limitations on conventional field monitoring efforts. This study presents an advanced IAE-YOLOv9 architecture, incorporating core modules—involution operator, alterable kernel convolution (AKConv), and element-wise multi-scale attention (EMA) mechanism—to enhance detection performance under complex field conditions. Notably, involution expands the effective receptive field, thereby strengthening detection of small-bodied targets. AKConv enables adaptive kernel reconfiguration, increasing responsiveness to scale variation. EMA facilitates refined object localization and classification by aggregating features across multiple spatial scales. When coupled with infrared-triggered camera traps, the system supports real-time surveillance, accurate detection, and automated counting of P. hodgsonii. On the self-constructed Sichuan-Gansu Wildlife Infrared Dataset (SGWID), IAE-YOLOv9 achieved 94.8% precision (95% confidence interval (CI): 94.2%–95.1%), 95.2% mean average precision (mAP@0.5) (95% CI: 94.4%–95.5%), and 91.2% recall (95% CI: 90.8%–91.5%), exceeding baseline detectors. Field deployment along migration corridors yielded 89.81% detection accuracy (95% CI: 88.34%–91.12%) and 88.28% counting accuracy (95% CI: 87.45%–89.12%) on the Hoh Xil Wildlife Camera Trap Dataset (HXWCTD), with high concordance to manual annotations. This integrative framework enables robust wildlife surveillance, migratory tracking, and conservation planning under high-altitude conditions, highlighting the effectiveness of combining infrared imaging with deep learning for ecological monitoring in extreme environments.
Rui et al. (Wed,) studied this question.