Land use has to be classified regularly. Now, thanks to a student at the Federal Institute of Technology in Lausanne (EPFL), the process has been further automated. According to a statement from the university, she has developed a machine learning algorithm that can distinguish between the forests that cover a country.
Valérie Zermatten trained her algorithm to recognize land categories such as rivers, lakes, campsites, sports fields, cemeteries, water purification stations, public parks, airports and dams. This makes it far superior to the Swiss Federal Statistical Office (FSO) solution called Arealstatistik Deep Learning (ADELE).
The results generated by Valérie Zermatten’s program are similar to the official data published by the FSO, indicating that it could be used in future land-use classification, writes the statement. One of its biggest advantages is in aerial photos, because classifying the images into some 40 different categories is still done mostly by hand. Switzerland now takes aerial photos of the land every three years, but the survey itself is only published every six years because classifying the images manually takes so much time. This type of mapping helps track urbanization, monitor soil permeability and combat urban sprawl.
“Our goal isn’t to replace humans with artificial intelligence,” said Devis Tuia, one of Zermatten’s thesis supervisors at the EPFL. “While Valérie’s algorithm will reduce the amount of painstaking work that must be done manually, human skills will still be needed for tasks that are beyond the scope of machines.” And there are plenty of human tasks, he added, such as distinguishing an apartment building from a school or a soccer field from a pasture.
