Home / Current Issue / Paper 1708844
Automatic Satellite Image Classification for Land Use and Land Cover Mapping Using Convolutional Neural Networks (CNNs)
Subject area: Science,Engineering and Technology · Area of research: Convolutional Neural Networks
Abstract
Using LULC, researchers can analyze the environment, plan cities, complete agricultural tasks and study climate change. Thanks to recent upgrades in satellites, it is now simpler to take good, crisp photos. Doing this for every image is a slow process, can involve our own prejudices and might not work for big projects. Using different images for training, CNNs from deep learning can determine the main features in a picture. The research paper outlines how satellite images for LULC can be classified using CNNs. During testing, Standard datasets EuroSAT and UCMerced were used along with custom-built and pre-trained CNNs, including ResNet50. I worked through the preprocessing steps, included new examples for data, trained the model, inspected accuracy, precision, recall, IoU and then compared their pictures. It seems, according to experiments, that models built using CNN are better and more accurate than traditional machine learning techniques. Even though the images differed little, both models managed to identify over 90% of each land cover type. Data imbalance, cloud effects and similar problems were overcome using augmentation and hyperparameter tuning. Actually, CNNs are dependable for mapping land use and land cover, making it easier to monitor the environment globally and apply the information quickly. The team aims to help edge devices benefit from machine learning and to make AI explanations available for their data teams.
Keywords
Land Use and Land Cover, Satellite Image Classification, Convolutional Neural Networks, Remote Sensing, Deep Learning, Semantic Segmentation, Environmental Monitoring, Supervised Learning, Earth Observation, Image Analysis
How to cite this paper
@article{1708844,
author = {Kamal Jyoti, Ayushi Upreti, Nitin Goel, Jubeen Misrty, Aniket Tripathi},
title = {Automatic Satellite Image Classification for Land Use and Land Cover Mapping Using Convolutional Neural Networks (CNNs)},
journal = {Iconic Research And Engineering Journals},
year = {2023},
volume = {7},
number = {6},
pages = {527-536},
issn = {2456-8880},
url = {https://www.irejournals.com/formatedpaper/1708844.pdf},
abstract = {Using LULC, researchers can analyze the environment, plan cities, complete agricultural tasks and study climate change. Thanks to recent upgrades in satellites, it is now simpler to take good, crisp photos. Doing this for every image is a slow process, can involve our own prejudices and might not work for big projects. Using different images for training, CNNs from deep learning can determine the main features in a picture. The research paper outlines how satellite images for LULC can be classified using CNNs. During testing, Standard datasets EuroSAT and UCMerced were used along with custom-built and pre-trained CNNs, including ResNet50. I worked through the preprocessing steps, included new examples for data, trained the model, inspected accuracy, precision, recall, IoU and then compared their pictures. It seems, according to experiments, that models built using CNN are better and more accurate than traditional machine learning techniques. Even though the images differed little, both models managed to identify over 90% of each land cover type. Data imbalance, cloud effects and similar problems were overcome using augmentation and hyperparameter tuning. Actually, CNNs are dependable for mapping land use and land cover, making it easier to monitor the environment globally and apply the information quickly. The team aims to help edge devices benefit from machine learning and to make AI explanations available for their data teams.},
keywords = {Land Use and Land Cover, Satellite Image Classification, Convolutional Neural Networks, Remote Sensing, Deep Learning, Semantic Segmentation, Environmental Monitoring, Supervised Learning, Earth Observation, Image Analysis },
month = {December},
}