RMS sometimes fails to identify bright meteors correctly. Why is this ?

To check for meteors, RMS looks for objects that move in a straight line through the video at a rate which is realistic for meteors. To do this, RMS identifies the centre of the potential meteor in each frame (known as the centroid).

Saturated Images

With most meteors the track is only a few pixels wide and a cross section through the track brightens then dims again in a classic bell curve around the central line, as shown below. This makes it fairly simple for RMS to calculate the location of the brightest point and hence calculate the track of the centroid.

However now consider a bright event like the one shown here. This has saturated the sensor over dozens or possibly hundreds of pixels to either side of the track. Effectively, the top of the peak gets "chopped off", and its harder to be sure about the location of the centroid. In reality the curve also won't be as symmetrical as this graph suggests because meteors do not burn up evenly, so the uncertainty in the location can be several tens of pixels.

It gets even worse if the meteor starts to fragment because as fragments separate from the main body, they create their own light curve which will overlap with the main body's light as shown simplistically here. Now, RMS will miscalculate the location of the centroid even more.

A few frames later, the fragment will have moved away (possibly creating its own track) and RMS will once again correctly calculate the centroid, but the calculated track is now "kinked" and RMS may reject the event as not "meteoric" enough.

Edge Effects

The problem gets even more interesting when a meteor passes out of shot at an angle to the edge of the field of view. When this happens, part of the light curve is lost offscreen, and the calculation of the centre is wrong. The track then takes on a 'hockey stick' shape, which can cause RMS to reject it as 'non-meteoric' and certainly means that the data are unreliable.

Internal Reflections

Extremely bright fireballs can give rise to internal reflections in the optics of the camera. These can move parallel to the true meteor or may be moving in the opposite direction. Either way they will completely confuse RMS's detector.

How do we fix this ?

The fix is to manually "reduce" the data as explained here

Impact on Magnitude Estimation

RMS estimates magnitude by calculating the brightness of the event and comparing it to the stars in the field of view. Brightness is the sum of the pixel values in the centroid, minus the average background sky brightness. For the vast majority of detections this works well. However as noted above, bright events can saturate the sensor. Obviously, if the top of the curve is 'chopped off', the brightness will be an underestimate, and therefore the magnitude will be an underestimate. Thus RMS routinely underestimates events that are bright as seen from a particular camera.

This highlights why its important to get multiple views on the same event: cameras further away won't suffer from sensor saturation and so will be more likely to accurately measure the brightness.

Can anything be done about it?

There are various mathematical techniques that can be used to estimate the "missing" part of the light curve. Application of these is an advanced technique and generally requires manual intervention.