I used fuzzywuzzy [1], a python-based fuzzy string matching calculator that is based on Levenshtein's edit-distance for a parking enforcement product I built.

I used fuzzywuzzy [1], a python-based fuzzy string matching calculator that is based on Levenshtein's edit-distance for a parking enforcement product I built. The product used an iOS client to capture license plates. The app would capture a single plate many times, de-duplicate using an edit-distance threshold matching plates up to a time period lookback. Then among those plates, send the one with the highest likely correct read up to the server. The web app would look to see if that plate had been seen before or a plate similar had been seen before. If so, it would join a "plate group" which let you see the history of vehicle sightings. A set of rules could be created to show vehicle in violation of a posted parking time limit. It worked when a person on patrol walked the lot twice and the app saw the "same" vehicle (thanks to edit distance) I had Cummings Properties in New England and National Cathedral in DC as customers, but sold the product last year. I did not like parking enforcement or the amount and type of data it handled. The product was hard to sell into facility management. Parking problems are a hot potato issue on private lots since they generally can’t ticket cars so it just takes time and money to deal with. That said the other solutions require manual entry or even clipboards. So people would get desperate to reduce effort involved. I really enjoyed building and iterating on the product, though, and learned a lot. I wrote some about Vladimir Losifovich Levenshtein in a final high level blog entry about edit distance. There's a 2002 photo of Levenshtein at a conference if you're curious. [2] [1] https://github.com/seatgeek/fuzzywuzzy https://github.com/seatgeek/fuzzywuzzy [2] https://blog.easyalpr.com/2019/09/16/calculating-license-plate-similarity-using-levenshtein-distance/ https://blog.easyalpr.com/2019/09/16/calculating-license-pla...