Journal of Earth System Science · 2024 · 19 citations · 77 references
ComputingComputer ScienceGeoinformaticsComputer VisionMachine LearningRemote SensingDeep LearningImage ClassificationLand Cover ClassificationLand Cover MappingRemote Sensing Image AnalysisRemote Sensing ImageryClassification Accuracy AssessmentConvolutional Neural NetworksLand Cover Classification SchemesPixel-based ClassificationPixel-wise ClassificationRemote Sensing ClassificationUrban Land Cover ClassificationUrban Land Cover Mapping
For proper planning of urban infrastructures such as road networks, pipelines, and other linear engineering structures, it is necessary to construct precise land use and land cover maps. Multiple attempts to develop land use land cover classification techniques have been made using various methods ranging from surveying to date image interpretation using remote sensing techniques. Land use land cover classification remains an intricate and challenging task due to the spectral and spatial complexity of the imagery. This work generates a labelled dataset called Sen-2 LULC ( https://data.mendeley.com/datasets/f4ky6ks248/3 ) using QGIS for land use land cover classification to train the state-of-the-art available convolutional neural networks with the backbone for pixel-wise classification of seven classes. This work deals with five models: UNet with ResNet50 backbone, UNet with ResNet152 backbone, UNet with DenseNet169, FPN with VGG16, and LinkNet with MobileNetv2. UNet-ResNet50 gives the best overall accuracy of 94.10% and Matthews correlation coefficient value of 0.69.
77
Deep Residual Learning for Image Recognition
Kaiming He, Xiangyu Zhang, Shaoqing Ren et al. · 2016 · 222.1K citations · Full text
Leo Breiman · Machine Learning · 2001 · 126.2K citations · Full text
Densely Connected Convolutional Networks
Gao Huang, Zhuang Liu, Laurens van der Maaten et al. · 2017 · 44.7K citations
Feature Pyramid Networks for Object Detection
Tsung-Yi Lin, Piotr Dollár, Ross Girshick et al. · 2017 · 28.7K citations