Classification and Object Detection on Satellite Images Using Custom CNN Architecture

Sunitha A N; Susmitha M N

doi:10.15680/IJIRSET.2023.1207190

Sunitha A N M.Tech Student, Department of Computer Science and Engineering, MVJ College of Engineering, Bengaluru, India
Susmitha M N Assistance Professor, Department of Electronics and Communication Engineering, MVJ College of Engineering, Bengaluru, India

DOI: https://doi.org/10.15680/IJIRSET.2023.1207190

Keywords: Image processing techniques include convolutional neural networks, satellite imaging, and object recognition and categorization.

Abstract

As the variety of applications for satellite image analysis growing, some of these applications include surveillance, military operations, geospatial research, and environmental impact and climate change monitoring. One of the most important aspects of satellite image analysis is the capacity to automatically recognise and classify objects. Recognizing and classifying items in aviation images may be challenging due to the type and size of the objects as well as the shifting visual features. Because of the type and information included in these photographs, manual article location is quite laborious. It is appealing to automate the finding of various highlights or articles from these satellite photos. The conventional techniques for categorising things include two steps: I identifying the areas of the image where objects are present, and (ii) categorising the items in those areas. It's harder to recognise items because of the complicated background, size, noise, and distance features. According to this research, a specially created convolutional neural network may be used to recognise and classify three distinct objects in the photographs, including trees, buildings, and cars. Also, it aims to grasp and briefly sum up the performance characteristics of the fictitious custom CNN.