Novelty detection in rover-based planetary surface images using autoencoders
In the domain of planetary science, novelty detection is gaining attention because of the operational opportunities it offers, including annotated data products and downlink prioritization. Using a variational autoencoder (VAE), this work improves upon state-of-the-art novelty detection performance...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2022-10-01
|
| Series: | Frontiers in Robotics and AI |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/articles/10.3389/frobt.2022.974397/full |
| _version_ | 1811338901244084224 |
|---|---|
| author | Braden Stefanuk Krzysztof Skonieczny |
| author_facet | Braden Stefanuk Krzysztof Skonieczny |
| author_sort | Braden Stefanuk |
| collection | DOAJ |
| description | In the domain of planetary science, novelty detection is gaining attention because of the operational opportunities it offers, including annotated data products and downlink prioritization. Using a variational autoencoder (VAE), this work improves upon state-of-the-art novelty detection performance in the context of Martian exploration by >7% (measured by the area under the receiver operating characteristic curve (ROC AUC)). Autoencoders, especially VAEs, perform well across all classes of novelties defined for Martian exploration. VAEs are shown to have high recall in the Martian context, making them particularly useful for on-ground processing. Convolutional autoencoders (CAEs), on the other hand, demonstrate high precision making them good candidates for onboard downlink prioritization. In our implementation adversarial autoencoders (AAEs) are also shown to perform on par with state-of-the-art. Dimensionality reduction is a key feature of autoencoders for novelty detection. In this study the impact of dimensionality reduction on detection quality is explored, showing that both VAEs and AAEs achieve comparable ROC AUCs to CAEs despite observably poorer (blurred) image reconstructions; this is observed both in Martian data and in lunar analogue data. |
| first_indexed | 2024-04-13T18:17:30Z |
| format | Article |
| id | doaj.art-f8d77bc3d8fd4f46b15a671042c4eb4f |
| institution | Directory Open Access Journal |
| issn | 2296-9144 |
| language | English |
| last_indexed | 2024-04-13T18:17:30Z |
| publishDate | 2022-10-01 |
| publisher | Frontiers Media S.A. |
| record_format | Article |
| series | Frontiers in Robotics and AI |
| spelling | doaj.art-f8d77bc3d8fd4f46b15a671042c4eb4f2022-12-22T02:35:37ZengFrontiers Media S.A.Frontiers in Robotics and AI2296-91442022-10-01910.3389/frobt.2022.974397974397Novelty detection in rover-based planetary surface images using autoencodersBraden StefanukKrzysztof SkoniecznyIn the domain of planetary science, novelty detection is gaining attention because of the operational opportunities it offers, including annotated data products and downlink prioritization. Using a variational autoencoder (VAE), this work improves upon state-of-the-art novelty detection performance in the context of Martian exploration by >7% (measured by the area under the receiver operating characteristic curve (ROC AUC)). Autoencoders, especially VAEs, perform well across all classes of novelties defined for Martian exploration. VAEs are shown to have high recall in the Martian context, making them particularly useful for on-ground processing. Convolutional autoencoders (CAEs), on the other hand, demonstrate high precision making them good candidates for onboard downlink prioritization. In our implementation adversarial autoencoders (AAEs) are also shown to perform on par with state-of-the-art. Dimensionality reduction is a key feature of autoencoders for novelty detection. In this study the impact of dimensionality reduction on detection quality is explored, showing that both VAEs and AAEs achieve comparable ROC AUCs to CAEs despite observably poorer (blurred) image reconstructions; this is observed both in Martian data and in lunar analogue data.https://www.frontiersin.org/articles/10.3389/frobt.2022.974397/fullautoencodernovelty detectionplanetary scienceplanetary roversprecision and recalldimensionality reduction |
| spellingShingle | Braden Stefanuk Krzysztof Skonieczny Novelty detection in rover-based planetary surface images using autoencoders Frontiers in Robotics and AI autoencoder novelty detection planetary science planetary rovers precision and recall dimensionality reduction |
| title | Novelty detection in rover-based planetary surface images using autoencoders |
| title_full | Novelty detection in rover-based planetary surface images using autoencoders |
| title_fullStr | Novelty detection in rover-based planetary surface images using autoencoders |
| title_full_unstemmed | Novelty detection in rover-based planetary surface images using autoencoders |
| title_short | Novelty detection in rover-based planetary surface images using autoencoders |
| title_sort | novelty detection in rover based planetary surface images using autoencoders |
| topic | autoencoder novelty detection planetary science planetary rovers precision and recall dimensionality reduction |
| url | https://www.frontiersin.org/articles/10.3389/frobt.2022.974397/full |
| work_keys_str_mv | AT bradenstefanuk noveltydetectioninroverbasedplanetarysurfaceimagesusingautoencoders AT krzysztofskonieczny noveltydetectioninroverbasedplanetarysurfaceimagesusingautoencoders |