High-Level Project Summary
The Safe Ground application aims to help prevent landslides and, consequently, develop communities through the integration of geographic data provided by specialized agencies and information provided by users. Therefore, Safe Ground aims to solve the problem through the use of machine learning techniques, such as neural networks and cross-correlation coefficients, producing models that allow for the prediction of locations with potential landslide risks, helping to avoid natural disasters and promote policies in partnership with government agencies.
Link to Project "Demo"
Link to Final Project
Detailed Project Description
The Safe Ground application provides its users with an analysis and mapping of risk areas in a more precise way, in addition, the application brings data of possible incidents in real time so that its users are alerted and move away from risk zones before for the event to happen. To make this possible, the application has machine learning techniques that are responsible for processing the data obtained via satellite and the data made available by our users.
We intend to use a supervised machine learning technique, an artificial neural network through the Backpropagation algorithm, in order to classify regions according to the degree of slip risk (low, medium and high), using the history of incidents obtained through the portal NASA Landslide Viewer.
In addition to this technique, we intend to use a data correlation technique and verify the existence of potential landslides in rural regions.
Using the history of incidents obtained through the NASA Landslide Viewer portal, it is possible to build a correlation of the data and verify the existence of potential landslides in rural regions.
For this correlation to be done assertively, we will use the technique devised by Zebende known as the Cross Correlation Coefficient.
This coefficient aims to quantify the level of correlation between non-stationary time series, based on the DFA (Detrended Fluctuation Analysis) and DDCA (Detrended Cross-Correlation Analysis) methods.
The cross-correlation coefficient ρDCCA is defined as the relationship between the covariance function without trend F2DCCA(n) and the variance function without trend FDFA(n) , ie:
Space Agency Data
All available links were used as a resource of the challenge to collect information, so we learned about the factors that impact the risks of landslides to happen and how this affects people's lives and the economy, being the basis of data from the Landslide Viewer portal and the Jaxa Alos satellite, the main ones, as the data provided by them are used in our project with the use of Machine Learning techniques (Neural networks and cross correlation coefficient) to be able to predict when a new case will happen sliding into a location more accurately.
Hackathon Journey
This journey helped us to understand how the geographic analysis of the planet is done through satellites, as well as to understand how to use such data in our tool. In addition, we had the opportunity to learn about a topic that directly affects communities in our country, which for us was very motivating. During the design of the tool, we planned activities based on the characteristics and knowledge of all team members, which helped and gave more dynamism during the elaboration and dissemination of ideas. In this way, the participation in the event was considered very enriching by the team, during the period in question, we further developed our ability to work collaboratively and solve problems.
We would like to thank the event organizers for the opportunity and learning, as we will carry with us throughout our career, we also thank Professor Leonardo Almeida for his mentorship and support during the conception of our ideas.
References
Leonardo Santana Almeida da Silva, Gilney Figueira Zebende. Estudo de Correlação Cruzada em Índices Econômicos. Disponível em: http://repositoriosenaiba.fieb.org.br/handle/fieb/763
Florêncio Mendes Oliveira Filho, Ivan Costa da Cunha Lima, Gilney Figueira Zebende. Amplitude da flutuação e correlação cruzada em sinais eletroencefalográficos: uma modelagem com a função de flutuação rms e o coeficiente de correlação cruzada ρDCCA. Disponível em: http://repositoriosenaiba.fieb.org.br/handle/fieb/1128
Aloísio Machado da Silva Filho, Gilney Figueira Zebende. Autocorrelação e Correlação Cruzada: Teorias e aplicações. Disponível em: http://repositoriosenaiba.fieb.org.br/handle/fieb/766
Nasa Landslide Viewer. Disponível em: https://maps.nccs.nasa.gov/arcgis/apps/webappviewer/index.html?id=824ea5864ec8423fb985b33ee6bc05b7
Jaxa Alos. Disponível em: https://www.eorc.jaxa.jp/ALOS/en/aw3d30/index.htm
Identifying Risk With Science + Communities Challenge Video. Disponível em:
https://www.youtube.com/watch?v=zamsH_btotY&list=PL37Yhb2zout05pUjr7OoRFpTNroq_wd9f&index=13
IMERG: Integrated Multi-satellitE Retrievals for GPM | NASA Global Precipitation Measurement Mission. Disponível em:
https://gpm.nasa.gov/data/imerg
Figma. Disponível em:
Gimp. Disponível em:
Kdenlive. Disponível em:
Instituto Brasileiro de Geografia e Estatística (IBGE). Disponível em:
Grossi, Enzo. Buscema Massimo. Introduction to articial neural networks. Disponível em: https://www.researchgate.net/publication/5847739_Introduction_to_artificial_neural_networks
Tags
#Science #Brazil #NASA #World
Global Judging
This project has been submitted for consideration during the Judging process.