Estimation of Remote Sensing Imagery Atmospheric Conditions Using Deep Learning and Image Classification
Estimation of atmospheric conditions is an important problem for remote sensing imagery analysis and processing. Especially it is useful to have a fast and accurate method when collecting weekly or daily imagery of the entire land surface of the earth with high resolution. This task appears in many remote sensing applications such as tracking changes of the landscape, agricultural image analysis, landscape anomaly detection. In this paper, we propose a method of atmospheric conditions estimation based on RGB image classification using fine-tunned CNN ensemble and image classifiers. We investigate usage of CNNs (Alexnet and a pretrained CNN ensemble) as feature extractors in combination with different classifiers such as XGBoost and ExtraTrees. We have tested the proposed method on a data set provided in the kaggle contest “Planet: Understanding the Amazon from Space” where the application task is to analyze deforestation in the Amazon Basin.
Korzh, Oxana and Serra, Edoardo. (2018). "Estimation of Remote Sensing Imagery Atmospheric Conditions Using Deep Learning and Image Classification". Intelligent Systems and Applications: Proceedings of the 2018 Intelligent Systems Conference (IntelliSys), 1237-1244. https://dx.doi.org/10.1007/978-3-030-01057-7_93