I got my research proposal accepted for the spring! I’ll be working on wetland mapping with deep learning techniques on high-resolution aerial imagery, and trying to accommodate spatial variability in geolocation compared with current one-size-fit-all neural networks. And I’m exicited to work with Professor Shashi Shekhar and his PhD student Jayant Gupta. I’d like to thank the Undergraduate Research Opportunities Program for providing this oppotunity.
As one of the largest ecosystems, wetlands have important value to the earth environment and human society. However, much value of wetlands has been reduced due to the degradation and loss of wetlands. Minnesota has lost about half of its original wetlands since 1850. Although historically wetlands were drained to support agricultural production and reduced for mining and road construction, in recent decades people have come to appreciate the many benefits they provide including flood reduction and water quality improvement. In order to provide information for the protection and restoration of wetlands, there is a need to update wetland inventory frequently with accurate boundary and improved delineation of smaller wetlands. As the Department of Natural Resources (DNR) claimed, the original National Wetland Inventory (NWI) was completed for Minnesota in the mid-1980s, and the recent statewide update took 9 years to complete (2010-2019), which is costly. So, we try to come up with a more efficient and economical way for future wetland mapping.