Dee Luo, 2016, Operations Research and Financial Engineering

The Mapping Africa project was formed as an initiative to collect a more accurate understanding of farmland distribution in Sub-Saharan Africa. Currently, the project uses crowd-sourcing to collect mapped data, which can be costly and inefficient. This summer, I worked with PhD student Stephanie Debats to develop a random forest algorithm to take satellite imagery of land in South Africa and accurately classify fields based on feature extraction. This type of land classification has not been explored much in the merging fields of remote sensing and computer vision. To accomplish this task, I learned about methods and applications of image segmentation, image processing techniques, and pixel-based feature classification. On the side, I also worked in R and QGIS to do spatial analysis on LANDSAT imagery as well as to build our available testing data for classification. Through this internship, I have been able to gain a stronger foundation in MATLAB programming and machine learning algorithms which I hope to continue to develop to help prepare me for a career in the technology industry, and to better understand and work toward solving global issues.