Fast Meteor Detection Toolbox - Archive ouverte HAL Access content directly
Poster Communications Year : 2022

Fast Meteor Detection Toolbox


Detection and characterization of meteoroids and space debris that enter into our atmosphere is one of the main concerns of astronomers. This work presents a computer vision application to detect meteors from video sequences. This application is designed to be embedded in satellites, in weather balloons or for airborne observation campaigns. The system runs on CPUs and is robust to non-stabilized cameras and noisy video sequences. It is evaluated on the 2022 Tau-Herculids video sequence where an overall tracking rate of 80.4% is obtained. All the visible meteors are detected. Experiment results demonstrate that the system runs efficiently with limited power thanks to the use of multithreading techniques (pipeline and fork-join parallelism). For example, on the 2022 Tau-Herculids HD video sequence, the system reaches 30 FPS on the Raspberry Pi 4 while the instant power is only 3.8 Watts.
Fichier principal
Vignette du fichier
poster_fmdt.pdf (5.48 Mo) Télécharger le fichier
XU4.png (1.15 Mo) Télécharger le fichier
fig_aff3ct_pipeline.pdf (102.08 Ko) Télécharger le fichier
fig_cluster_num.pdf (619.68 Ko) Télécharger le fichier
fig_detection_chain.pdf (57.56 Ko) Télécharger le fichier
fig_tasks_workload.pdf (47.18 Ko) Télécharger le fichier
logos.png (179.62 Ko) Télécharger le fichier
m1pro.jpg (468.9 Ko) Télécharger le fichier
meteorix_cao_upscale.png (1.24 Mo) Télécharger le fichier
nano.jpg (207.82 Ko) Télécharger le fichier
opensource.pdf (9.92 Ko) Télécharger le fichier
plot_meteors_duration_histo.pdf (24.84 Ko) Télécharger le fichier
rpi4.png (548.17 Ko) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

hal-03954992 , version 1 (26-01-2023)




Mathuran Kandeepan, Clara Ciocan, Maxime Millet, Manuel Bouyer, Adrien Cassagne, et al.. Fast Meteor Detection Toolbox. Workshop AFF3CT, Nov 2022, Bordeaux, France. , ⟨10.13140/RG.2.2.12222.36161⟩. ⟨hal-03954992⟩
80 View
38 Download



Gmail Mastodon Facebook X LinkedIn More