Research Interest
Deep learning approaches to building footprint extraction using high spatial resolution images
With the development of artificial intelligence, deep learning methods are widely used in remote sensing community. Using deep learning methods in high spatial resolution (HSR) images to extract buildings can effectively and efficiently provide buildings footprints which are useful for urban planning, high-definition mapping, change detection, and informed infrastructure management. However, pixel-based building extraction using HSR images brings three challenges: insufficient features, effects of shadows and occlusions, and massive data. For example, small buildings can be extracted accurately now, while large building cannot because more inter-class variations exist. Therefore, new AI-based methods should be investigated to effectively and efficiently extract building footprints in HSR images.
My current research focuses on characterization of different features, automated extraction of building footprints using deep neural networks, regularization building outlines using graphic models, development of novel building extractionalgorithms with improved accuracy, efficiency and robustness.
Education
- MSc, Cartography and Geographic Information Systems, Lanzhou University, 2019
- BSc, Geomatics, China University of Petroleum,Qingdao, 2016
Publications
- Wei, B.C.; Xie, Y.W.; Xu, J.; Wang, X.Y.; He, H.J.; Xue, X.Y., 2018. Land use/land cover change and its impacts on diurnal temperature range over the agricultural pastoral ecotone of Northern China, Land Degradation Development, 29, 30093020.