【Team】Geoinformatics Team
【Date】2026/October/15(Thursday) 13:00-14:00(JST)
【Speaker】Talk by Daoyuan Zheng, Wuhan University
Title: Weakly Supervised Remote Sensing Interpretation and Great Wall Heritage Monitoring
Abstract:
Accurate interpretation of high-resolution remote sensing imagery remains challenging because dense pixel-level annotations are costly to obtain, while many archaeological targets exhibit low contrast and partial occlusion caused by vegetation, terrain shadows, and complex backgrounds. This seminar presents two lines of research addressing these challenges. The first focuses on remote sensing interpretation under weak supervision, including building extraction and change detection using image-level and bounding-box annotations, with emphasis on recovering reliable spatial information from coarse supervision. The second focuses on remote sensing applications in Great Wall heritage monitoring, including terrain prior fusion for extracting weak linear Great Wall relics in mountainous regions, as well as automatic detection and site-scale vegetation shift analysis for individual heritage buildings along the Great Wall. These studies demonstrate a progression from learning with limited and coarse annotations to the interpretation and monitoring of challenging archaeological targets in real-world remote sensing applications.
