Finding Fireballs

From DFN Wiki
Jump to: navigation, search

This page will contain instructions for

  • recognising fireballs from satellites and other night-time phenomena
  • downloading and viewing images from DFN observatories
  • processing event detection emails
  • initiating the data processing pipeline for triangulation and orbits

Automated software for detecting fireballs

The images are automatically processed on the cameras.

The software basically works be differencing consecutive images, and tries to identify lines in the pictures.

That means that it is unable to tell apart fireballs from satellites or planes.

The software logs in a file called YYYY-MM-DD_DFNXXXNNN__log_processing.txt.

In addition, each detection generates several files:

  • .tile.jpg : light weight compressed JPG cut-out of the detection
  • .pixels.txt : rough pixel coordinates of the event
  • .raw_pixels.txt
  • .radec.txt : rough astrometric coordinates of the event

Central server vetting

As mentioned above, the results of the fireball detection software contain numerous false alerts (satellites, planes...), therefore the candidate events are vetted by a rough triangulation with other cameras.

Email alerts for human review

Emails are sent only when a triangulation is validated by the central server.

Valid detections

Each email is the result of a valid triangulation between 2 cameras detected by the server. This means that if more than 2 cameras detected the event, there are going to be one email for each pair of cameras.

Fireball. Typical (relatively faint) fireball. Shutter breaks are clearly visible.
Fireball. Very bright fireball observed from a large distance. At first look shutter breaks are hard to resolve, but this is indeed a fireball.

False positives

Satellite. The shutter breaks are hard to resolve because the apparent angular speed is much lower than that of a fireball.
Plane. Low flying planes have a similar angular speed as fireballs. It is impossible to tell them apart just by looking at a tile. Planes will typically be visible on several consecutive images, and can manoeuvre.