SOIL PERMEABILITY MAPS FROM VERY HIGH-RESOLUTION SATELLITE IMAGERY THROUGH FUZZY LOGIC AND OBJECT-BASED IMAGE ANALYSIS (OBIA)

The last ten years have seen a considerable increase in the use of satellite images for the purpose of terrestrial observation, at a technological level giant steps have been made, reaching very high ground resolution and considerably reducing the return times for the acquisition of the same area. This has opened up the use of satellite images in applications previously exclusive to aerial images, in fact at present for the most modern satellites it is possible to acquire images with a ground resolution of about 30 cm, being able to revisit the same area in a time between 2 and 5 days. The achievement of this technological milestone and the increase in demand has pushed more and more private entities to invest in the sector, creating entire constellations of satellites for commercial purposes, making the images increasingly accessible to the public also in economic terms. These images can be used for multiple purposes, ranging for example from monitoring large areas in situations of extreme emergency related to natural disasters, to planning the development of our cities, or in my case verifying the suitability of the images for the creation of soil permeability maps by automatically classifying the acquired area. The goal of these years, therefore, has been to develop a ground classification algorithm capable of achieving very high accuracy in the classification of the image of the municipal area of Pavia acquired by the WorldView3 satellite. This result was obtained by implementing a multilevel approach based on the analysis of the objects identified in the scene, applying for the most complex classes to be identified, a logical combination of different parameters based on the principles of fuzzy logic. The result was subsequently validated by comparing the classification obtained with the ground truth built by manually classifying approximately 10% of the analysed area. The overall accuracy achieved is equal to 95.72%, a decidedly satisfactory result considering that it was obtained for an area with a considerable extension and that it can be easily replicated on larger areas.