Applications of Convolutional Neural Networks to Landslides Detection on Multi-Spectral Image Time Series.

• Grant number: 25/22198-4
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: November 01, 2025
• End date: October 31, 2026
• Field of knowledge: Physical Sciences and Mathematics - Probability and Statistics - Statistics
• Principal Investigator: Aluísio de Souza Pinheiro
• Grantee: Francisco de Azevedo Patiño
• Host Institution: Instituto de Matemática, Estatística e Computação Científica (IMECC). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil
• Associated research grant: 23/02538-0 - Time series, wavelets, high dimensional data and applications, AP.TEM

Abstract

Landslides result in environmental and social setbacks. They are recurrent whenever one fiunds heavy rainfall and/or rought terrain. Particularly in Brazil we have faced extreme meteorological events. One such event happened in São Sebastião, SP, in February 2023, when a 600mmm precipitation in 24 hours resulted in dozen of deaths (Pivetta, 2023). Traditional mapping methods such as field mapping and visual detections are expenseive, slow and spatially limited (Guzzettti et al., 2012). Remote sensing provides high-resolution multi-temporal multi-spectral images.Deep learning, such as Convolutional Neural Networks (CNN's) and autoencoerds, revolutionized image analysis, outperforming more traditional machine learning methods (LeCun et al., 2015). CNNs automatically develop hierarchical image patterns, while Autoencoders learn anomalous behavior through compact representations (Zhang et al., 2025). The combination of these two paradigms may provide practitioners with precise automatic Earth's surface changes' detection.