Geographical Area Classification on Satellite Images Using CNN Architecture

Geographical Area Classification on Satellite Images Using CNN Architecture

Prof. Jagruti Wagh , Vivek Barhate, Preshita Bhortake, Ritika Kalotra

Computer Engineering Department MMCOE, Pune, India

Abstract—This paper presents a novel approach for classi- fying geographical areas in satellite imagery using a modified Convolutional Neural Network (CNN) architecture.The archi- tecture enhances feature extraction and classification accuracy by combining specialized layers like fully connected, pooling, and conventional layers. Our modified CNN performs better at accurately classifying a variety of geographic locations, according to our testing results. By efficiently capturing and analyzing complex spatial patterns, the use of customized layers enhances classification results in satellite-based geographical area classifi- cation.

Index Terms—Geographical area classification, Convolutional Neural Network (CNN), Satellite imagery, Image classification, Fully connected layers, Pooling layers, Conventional layers