Remote Sensing for Detection of Aviation Infrastructure and Ground Objects

Research Problem

High-spatial-resolution aerial and remote-sensing imagery provides valuable information for monitoring airports and other large infrastructure systems. However, aviation-relevant objects such as aircraft, vehicles, personnel, and small ground targets may occupy only a very small fraction of an image.

Detection is made more difficult by large image sizes, scale variation, occlusion, dense visual clutter, and real-time computational requirements. These challenges become particularly important for airport surveillance and future autonomous operations.

Contributions

Our research develops efficient deep-learning architectures for detecting small objects in aerial and remote-sensing imagery.

Key contributions include:

• Lightweight object-detection architectures designed for high-spatial-resolution imagery
• Normalization-free network designs that reduce computational complexity while maintaining detection performance
• Multi-scale feature extraction and feature-fusion methods for improving the detection of small and partially occluded objects
• Probabilistic bounding-box regression methods that improve localization of very small targets
• Evaluation using aerial and airport-surveillance datasets containing aircraft, vehicles, personnel, and other ground objects
• Development of methods relevant to aviation surveillance, infrastructure monitoring, and autonomous airport operations

Selected Publications

• A Lightweight Normalization-Free Architecture for Object Detection in High-Spatial-Resolution Remote Sensing Imagery
  Li, Y., Fang, Y., Zhou, S., Long, T., Zhang, Y., Ribeiro, N. A., & Melgani, F. (2025). IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 18, 24491–24508.

• Robust Small-Object Detection in Aerial Surveillance via Integrated Multi-Scale Probabilistic Framework
  Li, Y., Fang, Y., Zhou, S., Zhang, Y., & Ribeiro, N. A. (2025). Mathematics, 13(14), 2303.

International Collaborations

• Civil Aviation Flight University of China (CAFUC), China
  Collaboration on deep-learning methods for remote sensing and aerial airport surveillance.

Research Collaboration

• A*STAR Institute for Infocomm Research, Singapore
  Collaboration on efficient computer-vision methods for aerial and small-object detection.