Explainable AI for LULC Classification Using Sentinel-2 Satellite Imagery

Authors:

Mvn Mounika*¹, P Pranay Sai², G Geethika³, E Chaitra Teja⁴, Devara Nagasri⁵

^1,2,3,4Student, ⁵Professor

Department of CSE(AIML), Sreyas Institute of Engineering & Technology, Nagole Hyderabad, India.

E-Mail Id: amounik60@gmail.com*

ABSTRACT:

Earth depends heavily on how spaces get filled and managed. People's choices about ground usage matter just as much as the results they create. Spotting those patterns means studying images shot high above the world. Each photo check eats hours, drains budgets, slows progress. Tiny details pile up, turning clear views into tangled puzzles. Mistakes creep in faster when every pixel holds secrets hard to decode. This research built a method where machines inspect images to detect how terrain gets used. Not just guessing, the computer uses what's called a Convolutional Neural Network to scan scenes - finding woods, open plots, streams, settlements, or empty soil. While running, tweaks in design improved clarity in reading visual data. Often, artificial systems decide without explaining why - so here, researchers uncovered regions that drew focus when judgments formed. Watching highlights show up makes inner logic visible, peeling back steps hidden inside every output. A peek through that tiny frame turns foggy confusion into sharp understanding. Testing our approach across various image groups showed strong results, suggesting it could guide smarter choices on land use while supporting efforts to protect the planet.

Keywords: Satellite Image Analysis, Deep Learning, CNN, Remote Sensing, Explainable AI