講演抄録/キーワード |
講演名 |
2021-03-05 14:40
Retrieving Interesting Planetary Images based on Captions ○Hiya Roy・Toshihiko Yamasaki・Tatsuaki Hashimoto(UTokyo) PRMU2020-95 |
抄録 |
(和) |
(まだ登録されていません) |
(英) |
Planetary images are collected by sophisticated imaging devices onboard the orbiting or roving spacecraft. As the number of acquired images continues to grow exponentially with each ongoing mission, it becomes increasingly difficult to return all the data to Earth because of limited bandwidth between Earth and other planets and planetary bodies through Deep Space Network. In this paper, we aim to solve this problem by developing an onboard image analysis algorithm. We propose an image-text matching algorithm to enhance the onboard analysis capabilities by detecting objects in the observed image and sending only much smaller-sized metadata to Earth that describes the image. The proposed algorithm is also helpful to retrieve images of interest~(images with desired geologic and/or non-geologic features), from a large database based on query text. We use the Mars surface images captured by the navigation camera of NASA’s Mars Science Laboratory (MSL) Curiosity rover for experimental purposes and demonstrate that our method can accurately detect objects on the Mars surface images. Our experimental results also verify that our method can successfully retrieve images of interest, based on textual queries with superior performance (as evidenced by higher R@K value) compared to the state-of-the-art approaches. |
キーワード |
(和) |
/ / / / / / / |
(英) |
Image retrieval / Object detection / Mars surface images / / / / / |
文献情報 |
信学技報, vol. 120, no. 409, PRMU2020-95, pp. 151-156, 2021年3月. |
資料番号 |
PRMU2020-95 |
発行日 |
2021-02-25 (PRMU) |
ISSN |
Online edition: ISSN 2432-6380 |
著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します.(許諾番号:10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
PDFダウンロード |
PRMU2020-95 |