What would be the next frontier in AI for Earth Observation?
About Me
I am a second-year PhD student at the Computer Laboratory, University of Cambridge, supervised by Srinivasan Keshav, Andrew Blake, and David Coomes. I am also affiliated with the UKRI Centre for Doctoral Training in AI for Environmental Risks. My research lies at the intersection of computer vision, physics-informed machine learning, and climate change, developing a data driven system for scalable environmental monitoring.
Prior to Cambridge, I completed my Master’s at ETH Zurich in Switzerland with an Excellence Scholarship and my Bachelor’s at Nanjing University in China, both in Geomatics—an interdisciplinary field that has laid down the foundation for my work today, combining computer science with geospatial data.
Feel free to explore this website to learn more about my research portfolio and ongoing projects. I also share my recent research findings on LinkedIn. Beyond academics, I enjoy choral singing and Chinese calligraphy in my leisure time.
If you’re interested in my work or would like to connect, please feel free to reach out!
PhD Research Focus
My PhD research focuses on developing data-driven systems for scalable forest monitoring, integrating remote sensing and close-range observations.
On the remote sensing end, I leverage Earth Observation data by treating the radiative transfer model as a graphics renderer and its inversion as inverse graphics. Through an end-to-end machine learning pipeline, I retrieve interpretable biophysical variables from satellite imagery to characterize forests at a large scale.
On the close-range sensing end, I focus on training tree segmentation algorithms for 3D point clouds derived from drone-based laser scans. To overcome the challenges of limited annotated data, I generate synthetic datasets using graphics engines to train 3D vision algorithms for forest carbon quantification. The key challenge lies in bridging the domain gap between real and synthetic forests, ensuring robust algorithmic generalization.
Before completing my PhD, my goal is to develop a complete pipeline for an innovative forest monitoring system by connecting both ends—from remote to close-range sensing, from real to simulation. By inverting the radiative transfer process, we can extract biophysical forest parameters at scale. These variables, serving as meta-data to describe forest structure, will enable the creation of digital forests, facilitating the generation of synthetic data to train close-range sensing algorithms for precise forest monitoring anywhere on Earth.