Real-time Satellite Image Classification Using Convolutional Neural Networks for Earth Observation and Video Feed Analysis

Abstract

Satellite Image Classification is all about classifying satellite image datasets into their respective classes by assigning the correct class label. In supervised deep learning, we provide an efficient way of classifying earth observation imagery captured by satellites. By using Convolutional Neural Networks (CNN), we have achieved good accuracy as well as provided another way of testing the classifier by providing a video clip. This method has applications like live feeds from cameras or satellites, which can be processed frame by frame and tokenized, offering better accuracy rates. Furthermore, this approach enables real-time processing of satellite imagery for monitoring and analysis. Our model also demonstrates scalability, handling various satellite image datasets with diverse class numbers and dimensions. In this Satellite Image Classification paper, we have worked on different satellite image datasets, with varying class numbers and image dimensions. We have concluded results in this report, from which you will understand their details with respect to accuracy and performance.