9/5/2023 - Considerations for My Capstone Thesis in GIS
I started a Master's program in Geographic Information Systems last fall, which has taken up a fair amount of my free time (intentionally) and hindered
my blog. However, I have to brainstorm topics to do my capstone thesis on and what better place than here! My focus area is generally on geospatial
data and maybe app development, so I want to choose a more technical project.
Creating a dashboard or data pipeline would align with my focus area...but is admittedly boring unless I found a compelling topic. Instead, I want to follow up on a previous blog post where I discussed analyzing fictional maps as a part of literary analysis and understanding the 'space' that exists in a fictional setting. I think this would be a fascinating topic, but I think that I also did not focus on what is actually a pretty big challenge: digitizing the map.
Realistically, it makes the most sense to scan a hard copy of a book that I intend to use for its map. If it's a book that I want to do textual analysis with as well, I should also make sure I have a way to access a digitized copy of the text as well. When scanning the map, it is likely that it takes up multiple pages with a seam splitting the map or with the legend on a separate page from the map itself. Rarely, a map may even fold out beyond the typically boundary of the book. All of these will physically pose challenges; if I want to get a perfect scan, I may need to actually cut out pages and align them (though this feels like mutilating a book and the English major in me shivers).
Assuming I get a perfect scan of a map, there are further challenges: different map scales may be used, different cartographic styles may use different symbols for the same feature, fancy fonts or caligraphy makes text harder to read, non-point symbols may represent a point, in which case deciding exactly where it lies can be tricky at best, arbitrary at worst.
Despite these challenges, I think this would be a fun project to learn more about computer vision and see if I can create a model to identify common map features; at a minimum, the scale, map border, points of interest, and roadways would be ideal. I think that this may be tricky, but doable with some neural networks and training from traditional/historical maps. Taking this a step further, it would be great for my program to then format this information within a GIS so that an actual, interactive map file can be opened, essentially digitizing the map in a way beyond just scanning it. Such a file could be opened in ArcGIS Pro, uploaded as a web map, or for realistic fiction, added to Google Earth for viewing. This would also expand the ways in which I could compare the cartographic qualities of the maps with the literary ones of the text they reside in. Instead of just comparing scale to word count, I could track place names and visually display how much certain places are talked about on the map itself. Most of all, I could practice my programming skills and fulfill a requirement for my program at the same time!
I will probably revisit this idea after I finish A) Programming in GIS this fall where I'll learn more about how Python can be integrated with spatial information and B) Artificial Intelligence and Machine Learning for Geospatial Technology next spring, where I'll learn more conceptually about ML/AI and their capabilities.