Comparison of novelty detection methods for multispectral images in rover-based planetary exploration missions

Hannah R Kerner,Samantha Jacob,Heni Ben Amor,Danika F Wellington,James F Bell,Paul Horton,Brian D Bue,Kiri L Wagstaff,Chiman Kwan

doi:10.1007/s10618-020-00697-6

Abstract

Science teams for rover-based planetary exploration missions like the Mars Science Laboratory Curiosity rover have limited time for analyzing new data before making decisions about follow-up observations. There is a need for systems that can rapidly and intelligently extract information from planetary instrument datasets and focus attention on the most promising or novel observations. Several novelty detection methods have been explored in prior work for three-channel color images and non-image datasets, but few have considered multispectral or hyperspectral image datasets for the purpose of scientific discovery. We compared the performance of four novelty detection methods—Reed Xiaoli (RX) detectors, principal component analysis (PCA), autoencoders, and generative adversarial networks (GANs)—and the ability of each method to provide explanatory visualizations to help scientists understand and trust predictions made by the system. We show that pixel-wise RX and autoencoders trained with structural similarity (SSIM) loss can detect morphological novelties that are not detected by PCA, GANs, and mean squared error autoencoders, but that the latter methods are better suited for detecting spectral novelties—i.e., the best method for a given setting depends on the type of novelties that are sought. Additionally, we find that autoencoders provide the most useful explanatory visualizations for enabling users to understand and trust model detections, and that existing GAN approaches to novelty detection may be limited in this respect.

Highlights

The goal of novelty detection approaches is to identify patterns in data that have not been previously observed (Markou and Singh 2003a, b; Chandola et al 2009; Pimentel et al 2014)
We compared the performance of autoencoder, Generative adversarial networks (GANs), principal component analysis (PCA), and RX approaches for prioritizing images with novel geologic features in multispectral images of the Martian surface acquired by the Mastcam imaging system in order to accelerate tactical planning for the Mars Science Laboratory (MSL) Curiosity rover
We found that the RX method had the best overall performance as measured by receiver operating characteristics (ROC) area under the curve (AUC) score and Precision @ N, but may not provide the most effective explanatory visualizations for allowing users to understand the features in an image that were detected as novel

Summary

Introduction

The goal of novelty detection approaches is to identify patterns in data that have not been previously observed (Markou and Singh 2003a, b; Chandola et al 2009; Pimentel et al 2014). Successive rover drives makes follow-up observation of late-identified science targets increasingly more costly to mission resources to pursue (since the rover would need to reverse course to re-visit the target). These factors require scientists to review the latest science data and identify targets of interest for follow-up analysis in a relatively short amount of time. There is a need for systems that can rapidly and intelligently extract information of interest from science instrument data to focus on potential discoveries and avoid missed science opportunities These systems must provide explanatory visualizations that allow scientists to trust and understand how a system came to its conclusion, a need that has not been explored extensively in prior work. This work aims to enable planning and data analysis teams to spend their limited available time on the most promising, or novel, observations

Objectives

Methods

Discussion

Conclusion