1. Meteor detection software for low-cost single-board computers Denis Vida Dario Zubović, Peter Gural, Damir Šegon, Robert Cupec

2. Recap - hardware Sony Effio ICX 673 camera 4mm f/1.2 EasyCap UTV007 digitizer RaspberryPi 2 single-board computer 4 cores @ 1GHz (OC), 1GB RAM, 64GB SD card 10-20x slower than PCs today TOTAL: ~$150 2 Vida et al., Meteor detection software for low-cost single-board computers

3. Recap - software Real-time: Video capture Compression to CAMS FTP format (FF files) Fireball detection Open source: Python, C++ – Available on GitHub Meteor detection Calibration procedure 3 https://github.com/CroatianMeteorNetwork Vida et al., Meteor detection software for low-cost single-board computers

4. Current system overview CAMS compatible Individual modules tested System testing 4 Vida et al., Meteor detection software for low-cost single-board computers

5. Requirements Max ~30s processing time per FF file on Raspberry Pi 2 Reliable star detection  Calibration – Rejecting images with no stars  Faster processing – High accuracy star positions  Better astrometry Robust meteor detection – High enough detection rate to be competitve – Low false positive rate  Less time in postprocessing 5 Vida et al., Meteor detection software for low-cost single-board computers

6. Star extraction Local maxima detection  Candidate stars Fitting a PSF to all candidate stars – Rejecting non-plausible candidates by evaluating the PSF covariance matrix (i.e. star’s shape) 6 Vida et al., Meteor detection software for low-cost single-board computers

7. Star extractor performance Average ~3s execution time per FF file on RPi2 Average of 30 stars on image – Average cumulative of 40 000+ stars per night PSF fitting robustly discriminates stars from noise – Only ~5% noise Detects faint stars VERY precise star positions due to PSF fitting Intensity calculated from the PSF 7 Vida et al., Meteor detection software for low-cost single-board computers

8. Meteor detector Runs ONLY when there are stars present on the image 8 CAMS FTP image format: Raw maxpixel image Vida et al., Meteor detection software for low-cost single-board computers

9. Meteor detector Step 1 – Thresholding FF file: max – avg > 1.7 * stddev + 9 9 Thresholded image Vida et al., Meteor detection software for low-cost single-board computers

10. Meteor detector Step 2 – Apply morphological operations  Meteors = lines 10 Morphed image Vida et al., Meteor detection software for low-cost single-board computers

11. Meteor detector Step 3 – Kernel-Based Hough Transform (open source!) 11 Identified lines Vida et al., Meteor detection software for low-cost single-board computers

12. Meteor detector Step 4 – Check for temporal propagation & merge similar lines 12 Frame X axis Y axis Vida et al., Meteor detection software for low-cost single-board computers

13. Meteor detector Step 5 – Centroiding 13 Vida et al., Meteor detection software for low-cost single-board computers

14. Meteor detector performance Evaluated on 100 carefully chosen meteor images – Wide variety of brightness, duration and velocity Evaluated on 5 full nights (over 3000 images each) – Mixed sky conditions (clear, overcast, rain, storm, full moon) Average ~9.5s execution time per FF file on RPi2 Detection rate similar to MeteorScan (on CMN data) False positive rate considerably smaller than MeteorScan Rejects meteors shorter than 4 frames in duration More tests to be done… 14 Vida et al., Meteor detection software for low-cost single-board computers

15. Conclusion A fully operational low-cost meteor station in sight! To be done soon: – Calibration procedure (astrometry done, photometry soon) – System test Future improvements: – Better meteor detector? – Make it more user friendly Why should YOU be interested? – Eductional tool - one meteor station per school? – Cheap way to expand or create new meteor networks 15 Vida et al., Meteor detection software for low-cost single-board computers

16. Thank you for your attention! Questions?