Skip to content

SatRed: New classification land use/land cover model based on multi-spectral satellite images and neural networks applied to a semiarid valley of Patagonia (2022)

Notifications You must be signed in to change notification settings

aletrujim/SatRed

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

76 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

SatRed: New classification land use/land cover model based on multi-spectral satellite images and neural networks applied to a semiarid valley of Patagonia

satred

Tabla de Contenido
  1. About The Project
  2. Getting Started
  3. Results
  4. Citation
  5. Contact

About The Project

satred

In this article we describe a new model, SatRed, which classifies land use and land cover (LULC) from Sentinel-2 imagery and data acquired in the field. SatRed performs pixel-level classification and is based on a densely-connected neural network. The study site is the lower Chubut river valley which has an extension of 225 km2 and is located in estern semiarid Patagonia. SatRed showed a 0.909 ± 0.009% (mean ± sd, n = 7) overall accuracy and outperformed the seven most traditional Machine Learning methods, including Random Forest. Our model accurately predicted buildings, shrublands, pastures and water and yielded the best results with classes harder to classify by all methods considered (Fruit crops and Horticulture). Further improvements involving textural information and multi-temporal images are proposed. Our model proved to be easy to run and use, fast to execute and flexible. We highlight the capacity of SatRed to classify LULC in small study areas as compared to large data sets usually needed for state-of-the-art Deep Learning models suggested in literature.

Getting Started

All procedures were computed in a virtual machine sponsored by Microsoft Azure 2, with the operating system Windows Server 2016 Datacenter, of standard NC6 size (6 vCPUs [Intel Xeon CPU E5-2690 v3 2.60 GHz], 56 GB RAM memory and a NVIDIA Tesla K80 GPU coprocessor). Algorithms were implemented using the Python programming language. The Rasterio package (Gillies et al., 2013) was used to access and process geospatial raster data, and Shapely (Gillies et al., 2019) for manipulation of polygons.

To run a local copy of this project follow these simple steps:

Prerequisites

  • GDAL
    pip install GDAL
  • Tensorflow
    pip install --upgrade tensorflow
  • scikit-learn
    pip install scikit-learn

Installation

  1. Download training and validation images Link or prepare your own. If you use the downloaded images, please note that they were prepared and partitioned with Partition
  2. Clone the repo
    git clone https://github.com/aletrujim/SatRed.git
  3. Install packages
    pip install -r requirements.txt
  4. Run python script
    python compare-classifiers.py --train=train --test=test --segmented=result
    python satred.py --train=train --test=test --segmented=result --epochs=250

Results

satred results

We developed a Neural Networks-based model for pixel-based classification of satellite imagery to map land use and land cover, including crops. SatRed has a relatively low requirement of data compared to state-of-the-art deep learning models suggested by the literature. In addition, SatRed has an overall good performance and outperforms other seven machine learning methods by at least 3.5% in overall precision scores. It also stands out for being more stable in individual classes, showing better scores for all classes being analysed. SatRed training is also faster than Nearest Neighbors which also shows acceptable scores.

Citation

If you use this data or the SatRed model in your research, please cite this project.

@article{trujillo2022satred,
  title={SatRed: New classification land use/land cover model based on multi-spectral satellite images and neural networks applied to a semiarid valley of Patagonia},
  author={Trujillo-Jim{\'e}nez, Magda Alexandra and Liberoff, Ana Laura and Pessacg, Natalia and Pacheco, Cristian and D{\'\i}az, Lucas and Flaherty, Silvia},
  journal={Remote Sensing Applications: Society and Environment},
  volume={26},
  pages={100703},
  year={2022},
  publisher={Elsevier}
}

Contact

Alexa Trujillo - @aletrujim

Project Link: https://github.com/aletrujim/SatRed

About

SatRed: New classification land use/land cover model based on multi-spectral satellite images and neural networks applied to a semiarid valley of Patagonia (2022)

Resources

Stars

Watchers

Forks

Packages

No packages published

Languages