Asteroid Zoo Talk

Classifying asteroid observations

  • grums by grums

    I note that we all have different criteria for deciding whether to mark a pattern of marks on the image sets an “asteroid” or to grade it as “nothing”. There is no specific guidance on whether to be generous or severe in making such judgements and, without understanding how these assessments are followed up it is hard to know which way to lean. I guess it is done by only following up image sets that have several people identifying the same asteroid. This seems a reasonable approach but may not be ideal in terms of making the most efficient use of the “crowd”. I think many people spend a long time identifying what turn out to be random noise sequences. I know, because I have done this myself. If you come back to the image set later you may find an entirely different one and may not recognise the one you picked earlier. The human brain is particularly good and finding a pattern, even where there isn’t one – remember the “canals” on Mars. Spending less time searching image sets with a lot of “noise” may allow more image sets to be processed and would give a higher hit rate.

    As we would all make different assessments I suggest that it may be useful to use a consistent method of grading the quality of a potential asteroid track. It could be just used in the comments (talk) to see whether it works or not. The criteria for a track to be an asteroid, in every case, must include the following:-

    The sequence of points in frames 1 to 4 (if visible in all frames) must be in a straight line and be equally spaced. Allowance can be made for frame to frame jitter and some “spreading” of points due to atmospheric, or other disturbance, and image filtering.

    Other than that being an essential requirement here is my assessment of how to grade the likelihood of the observation actually being an asteroid:

    A. 4 bright roughly round marks (each comprising several pixels)

    B. 3 bright roughly round marks (each comprising several pixels). One not visible because of the track extending over the frame edge, because of bad set affecting one frame or because of excessive jitter.

    C. 2 bright roughly round marks (each comprising several pixels). As for B but an extra qualification that the brightness and size of the two marks must be high compared to the brightest background noise i.e. similar to a star. The spacing of the two marks must be less than 5% of the image width.

    D. 4 marks comprising several pixels with brightness at a similar level to the brightest background noise. The marks may be variable in brightness, shape and size due to atmospheric distortion.

    E. 4 marks comprising several pixels at a level comparable to that of the typical background noise. The marks may be variable in brightness, shape and size due to atmospheric distortion. There should be no contra-indications (see below).

    F. As for E but where one of the marks, although visible, is at a level that could not be distinguished from background fluctuations. There should be no contra-indications (see below).


    1. Look at the local, surrounding background noise. If it has a tendency to be “moving” in the same direction and speed as the potential asteroid sequence, it probably means that the sequence is resulting from atmospheric effects (e.g. moving thin clouds occluding background stars). There may be other reasons but it is clear that it could be a likely cause of the observed sequence.
    2. If there are several such sequences in the set it may mean that there is generally too much noise (maybe due to clouds and over-exposure). It throws too much doubt on their validity.
    3. If one of the marks in the sequence re-appears in in the same place in another frame. This may well indicate that the sequence is due to atmospheric movement occluding a fixed object rather than an object moving.

    To describe an asteroid track position, for others to look at, it is important to note that the image on your screen may not be the same size on other’s screens so it is no good using centimetres. I tend to use a clock face to indicate its position in frame 1 as an angle from the centre and a distance from an edge as a percentage of the image width or height. So, for example: just after 2 oclock 16% image width from RHS
    This is just an idea and I will try using this in “Talk” to see how it goes. Any comments welcome.