タイトル（掲載誌）宇宙航空研究開発機構特別資料: 第9回スペースデブリワークショップ講演資料集 = JAXA Special Publication: Proceedings of the 9th Space Debris Workshop
一般注記第9回スペースデブリワークショップ (2021年2月24-26日. 宇宙航空研究開発機構調布航空宇宙センター(JAXA)(CAC), オンライン会議), 調布市, 東京
The 9th Space Debris Workshop (February 24-26, 2021. Chofu Aerospace Center, Japan Aerospace Exploration Agency (JAXA)(CAC), Online meeting), Chofu, Tokyo, Japan
We have developed a streaks detection algorithm for images coming from the Tomo-e camera of the 1M telescope at Kiso Observatory. It follows a computer vision approach, with the next stages: preprocessing, thresholding, Hough transform, and clustering of streaks. A set of artificial streaks were superimposed over real Tomo-e images, and with a background noise of the real frames estimated in 550 ADU, and a peak brightness of the artificial streaks ranging from 630 to 1150 ADU, all streaks were detected correctly. The algorithm took 0.52 seconds to detect 8 artificial streaks in one of the 84 image sensors, when running in a 1.1GHz dual core CPU processor with RAM 8 GB. Real images will be tested at a later stage. The Tomo-e camera survey has a minimum time between frames of 6 seconds, which corresponds to 12 consecutive images of 0.5 seconds exposure each, for one particular region of the sky. Currently the algorithm is being exported to a GPU NVIDIA Quadro RTX 8000 dual system with 144 multiprocessors, that will allow the algorithm to process in parallel over the 84 image sensors of the camera, with the aim to detect all streaks under 6 seconds.
形態: カラー図版あり
Physical characteristics: Original contains color illustrations
資料番号: AA2130003047
レポート番号: JAXA-SP-21-001
連携機関・データベース国立情報学研究所 : 学術機関リポジトリデータベース（IRDB）（機関リポジトリ）
提供元機関・データベース宇宙航空研究開発機構 : 宇宙航空研究開発機構リポジトリ