Abstract
Disaster relief, police work, and environmental checks rely heavily on
satellite imagery. Some users need human assistance manually identifying
facilities and items in the photos. Due to the significant regions that
need to be searched and the scarcity of available analysts, automation
is crucial. Yet, standard object identification and classification
techniques must improve accuracy to solve the problem. A set of machine
learning techniques called “deep learning” has shown promise for
automating these operations. It has had success using convolutional
neural networks to comprehend images. Using high-resolution,
multi-spectral satellite images, we apply them to the issue of object
and facility identification in this work. We outline a deep-learning
object classification system. We use the Satellite Image Classification
Dataset-RSI- CB256 for this study. This dataset has four classifications
combined using Google Maps snapshots and sensors. Suggested hybrid model
in this paper gives an accuracy of 98.96%.