Machine Learning Pipeline for Shift-Invariant Detection of Volcanoes on Venus

Abstract

Intelligent algorithms are constantly being developed to improve the ability of machines to extract and process meaningful data in a variety of situations. In this work, we present a machine learning pipeline that streamlines the task of selecting preprocessing algorithms, feature extraction algorithms, and classification algorithms. We demonstrate the pipeline by identifying volcanoes in synthetic aperture radar (SAR) images of the surface of the planet Venus. This dataset is imbalanced, in the sense that there are relatively few images containing volcanoes, which is a common situation in many autonomous sensing tasks. We show that our machine learning pipeline is able to identify a set of algorithms that can be used together to identify volcanoes with high recall. While the precision of the classifier is poor, it can still be used to reduce the overall size of the dataset and improve the balance of the dataset.