Mapping Purple Loosestrife at Lake Neshonoc, WI Using Deep Learning Classification Technique on Drone Imagery

Authors: Ethan Lucas*, University of Wisconsin - La Crosse, Gargi Chaudhuri, Professor and advisor, Niti Mishra, Professor and advisor
Topics: Remote Sensing, Geographic Information Science and Systems, Environment
Keywords: Deep Learning, Purple Loosestrife, Image Classification, Pixel Based
Session Type: Poster
Presentation File: No File Uploaded


The aim of this research is to create a specie-level map of Lake Neshonoc area using deep learning based pixel –level image classification on very high resolution (VHR) images acquired by Unmanned Aerial System (UAS, popularly called drones). The target of this classification is Lythrum salicaria or purple loosestrife. Purple loosestrife is an invasive species that has made its home in many of the wetlands in Wisconsin, and it is important to monitor and if needed, control this plant. Purple loosestrife can grow in extremely dense clumps and push out native plant species in the process. This has the potential to lower the biodiversity in the area. It also can take over the habitats of animals that lived there, further restricting available habitable space for some species of animals. These dense clumps can also make recreational activities difficult, taking away from the local economies of the areas affected (Minnesota Department of Natural Resources). The goal of this research is to produce an accurate map of the purple loosestrife which will be useful to take necessary actions to curb further growth of invasive species. This could also provide vital information on whether efforts of reducing this invasive species are effective or not. With the increasing availability of artificial intelligence (AI) and deep learning, application of these technologies to develop detailed and accurate land cover map will be highly beneficial for conservation efforts.

To access contact information login