Analyse spatio-temporelle de signaux pour détecter la dérivede capteurs mesurant la qualité de l’air
Detection of drift concept in a sensor network from the Graph Fourier Transform (GFT).
Explore the docs »
View Demo
·
Report Bug
·
Request Feature
The network of sensors allows to follow a complex phenomenon by observing the temporal evolution of the values in several points, and by crossing the information from one sensor to another. This allows, for example, to carry out a meteorological or seismic monitoring, by detecting a cloud or a tremor. However, the sensors used are sometimes subject to drift, wear and tear or possible interference, which makes some of the observed values false. It is therefore important to be able to determine if a sensor starts to drift in order to allow a good analysis of the data. Therefore, the objective of this project is to detect a drifting sensor (when and where) in a real sensor's network. For this, we will explore the usefulness of a particular mathematical tool: the graph space-time spectrogram. Such a tool allows to decompose a space-time series into a sum of space-time frequencies, in a similar way to Fourier analysis, but for signals in graphs, known as Graph Fourier Transform (GFT).
To get a local copy up and running follow these simple steps.
It's possible that need install the following code
sudo apt-get install python3-dev graphviz libgraphviz-dev pkg-config
- Clone the repo
git clone https://github.com/Johansmm/air-polution-sensor
- Install requerements
python3 -m pip install -U requerements
Run the different notebooks that is contained in this repository. The main objectives of the project were developed in the sensors_drift.ipynb notebook. However, you can use the following notebooks:
examples/First_spectogram_tests.ipynb: Manipulation of pygsp library. Acknowledgments to BastienPasdeloup.examples/GFT_sensors.ipynb: Presentation of a network and synthetic signals in order to visualize some concepts such as GFT, spectrograms and drift concept.examples/Transdim.ipynb: Data Imputation using the library transdim , library developed in MIT to handle missing values in spatio tamporal data modeling.
See the open issues for a list of proposed features (and known issues).
Contributions are what make the open source community such an amazing place to be learn, inspire, and create. Any contributions you make are greatly appreciated.
- Fork the Project
- Create your Feature Branch (
git checkout -b feature/AmazingFeature) - Commit your Changes (
git commit -m 'Add some AmazingFeature') - Push to the Branch (
git push origin feature/AmazingFeature) - Open a Pull Request
Distributed under the MIT License. See LICENSE for more information.
- Johan Mejia (johan-steven.mejia-mogollon@imt-atlantique.net) -
- Tatiana Moreno (jenny-tatiana.moreno-perea@imt-atlantique.net) -
- Diego Carreño (diego-andres.carreno-avila@imt-atlantique.net) -
- Ilias Amal (ilias.amal@imt-atlantique.net) -
- Project Link: https://github.com/Johansmm/air-polution-sensor
The authors would like to thank AlexandreReiffers and BastienPasdeloup for their support and follow-up during the development of the project.
We also thank Best-README-Template for providing the repository documentation.