TY - JOUR

T1 - Unsing Time Series Image Data To Improve The Generalization Capabilities Of A CNN

T2 - ISPRS Geospatial Week 2023

AU - Ortega, M. X.

AU - Wittich, D.

AU - Rottensteiner, F.

AU - Heipke, C.

AU - Feitosa, R. Q.

N1 - Funding Information: This work was conducted in the scope of ForstCARe project, supported by and the Federal Ministry for Economic Affairs and Climate Action, Germany (Bundesministerium fur Wirtschaft und Klimaschutz, Funding codes 50EE2017A and 50EE2017B). The authors would also like to acknowledge the support provided by the DAAD (Deutscher Akademischer Austauschdienst) through the funding program 57588368.

PY - 2023/12/5

Y1 - 2023/12/5

N2 - Deforestation is considered one of the main causes of global warming and biodiversity reduction. Therefore, early detection of deforestation processes is of paramount importance to preserve environmental resources. Currently, there is plenty of research focused on detecting deforestation from satellite imagery using Convolutional Neural Networks (CNNs). Although these works yield remarkable results, most of them employ pairs of images and detect changes which occurred between the two image acquisition epochs only. Furthermore, these models tend to produce poor results when applied to new data in real-world scenarios. In this regard, an interesting research topic deals with the generalization capacity of the classifiers. CNN-based approaches combined with time series data can be a suitable framework to obtain classifiers that generalize better to new data. Image time series contain complementary information, representing different imaging conditions over time. This work addresses the transferability for detecting deforestation in different areas of the Amazon region, using Sentinel-2 time series and reference maps from PRODES project, which are not required to be synchronized. The results indicate that the classifier with time series data brings a substantial improvement in accuracy by taking advantage of the temporal information.

AB - Deforestation is considered one of the main causes of global warming and biodiversity reduction. Therefore, early detection of deforestation processes is of paramount importance to preserve environmental resources. Currently, there is plenty of research focused on detecting deforestation from satellite imagery using Convolutional Neural Networks (CNNs). Although these works yield remarkable results, most of them employ pairs of images and detect changes which occurred between the two image acquisition epochs only. Furthermore, these models tend to produce poor results when applied to new data in real-world scenarios. In this regard, an interesting research topic deals with the generalization capacity of the classifiers. CNN-based approaches combined with time series data can be a suitable framework to obtain classifiers that generalize better to new data. Image time series contain complementary information, representing different imaging conditions over time. This work addresses the transferability for detecting deforestation in different areas of the Amazon region, using Sentinel-2 time series and reference maps from PRODES project, which are not required to be synchronized. The results indicate that the classifier with time series data brings a substantial improvement in accuracy by taking advantage of the temporal information.

KW - Convolutional neural networks

KW - Deforestation detection

KW - Time series

KW - Transferability

UR - http://www.scopus.com/inward/record.url?scp=85182995319&partnerID=8YFLogxK

U2 - 10.5194/isprs-annals-X-1-W1-2023-961-2023

DO - 10.5194/isprs-annals-X-1-W1-2023-961-2023

M3 - Conference article

AN - SCOPUS:85182995319

VL - 10

SP - 961

EP - 970

JO - ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences

JF - ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences

SN - 2194-9042

IS - 1

Y2 - 2 September 2023 through 7 September 2023

ER -