Re: Satellite Elements from Astrometric Observation.

From: John Burns (
Date: Tue Mar 17 2009 - 12:43:14 UTC

  • Next message: Ted Molczan: "Updated elements"

    You've pretty much hit it on the nose with what I was thinking.
    I've created some image detection code which already detects bright  
    points as an X,Y and brightness.  I'm then iterating through every  
    possible set of 3 bright points in the image (brightest first) and  
    calculating the angle and (relative) length of the sides for comparison.
    Other problems may arise such as lens distortion, however this should  
    be minimal at fields of view narrower than say 30 degrees.
    I've got two pictures at the moment which are suitable for development  
    but I'll definitely get back in touch once the code is ready for  
    testing against other images.
    On 17 Mar 2009, at 12:19, Greg Roberts wrote:
    > Afternoon John
    > I could supply you with some of the images I have taken - I would  
    > rate them as fair quality for several reasons. I can provide the  
    > approximate centre of the field of view as well as field of view.  
    > Are you okay with FITS images- this has the time of the end of  
    > exposure in the header file.
    > Now normally my images are taken with the camera stationary which  
    > means the geostationary satellite will appear stellar and the stars  
    > trail with the left edge of the star trail corresponding to the time  
    > the exposure was recorded. The magnitude limit will typically be  
    > around 12-14th magnitude.
    > I could also supply some FITS images where the camera is tracking at  
    > sidereal rate so stars appear stellar-like but there will be some  
    > edge distortion and field rotation as Im using an altazimuth  
    > mounting . I do NOT normally use this mode for satellites since my  
    > geostats will then trail and I loose sensitivity - however this is  
    > probably the type of picture that video observers will take of leo  
    > satellites ...
    > Anyway if of any interest please let me know and Ill be happy to  
    > send. I could of course send the images as JPEG files and identify  
    > field and time of exposure in the naming of the file.
    > As far as satellite elements from astrometric observation - (not  
    > what you are offering to do) but since I "have the floor" Id like to  
    > say how I would possibly like it to work:
    > One takes a picture - FITS or JPEG or extracts it from an avi file  
    > so you have the time and the satellite name. One also has an idea of  
    > how close the object was to prediction in time and position so one  
    > starts with a list of satellites and corresponding time of  
    > observation and approximate error in time and position ( say to 2-3  
    > secs and position say 0.5 degrees) as well as the elements used for  
    > the observation as well as the matching images.
    > Now it would be NICE if this is all one has to do- the program reads  
    > the satellite name and time and program computes the expected  
    > position. Now comes the tricky part- the program needs to identify  
    > this position against the stellar background in the image so the  
    > program needs access to a database to at least mag 12 stars. The  
    > program will then try and pattern match the expected star background  
    > against the observed background, starting with the predicted  
    > position and, if not matched, then correct the predicted position by  
    > the estimated time difference noted and/or positional uncertainty,  
    > so means sliding the obtained image along the new predicted  
    > satellite track and finding a match. This should not be too hard a  
    > job for todays pc's if the images were stellar but they will not be  
    > - at least in my case, so it means "plotting" the end of each star  
    > trail to get the stars x,y coordinates. Then the program also needs  
    > to be able to identify which is the "intruder" - ie the satellite  
    > and, if the satellite is trailed which is the end you want to measure.
    > It will also probably to necessary to be able to rotate the expected  
    > star field to match the orientation of the image which may have been  
    > taken with an altaz /equatorial or other type of mounting. The  
    > observer should usually be able to supply a reasonably good estimate  
    > of the field of view which will narrow the pattern matching somewhat.
    > Once one has the RA/Dec of the satellite the orbit reduction is  
    > pretty straightforward as such programs already exist and in some  
    > cases the code is available - eg Scotts software for example. It is  
    > also possible to get a satellite magnitude from the intensity but I  
    > would not put too much faith in this ...
    > I think the entire procedure is pretty straight forward but I think  
    > its beyond my programming capabilities .
    > Anyway just some ideas...
    > Cheers
    > Greg
    Subscribe/Unsubscribe info, Frequently Asked Questions, SeeSat-L archive:

    This archive was generated by hypermail 2b29 : Tue Mar 17 2009 - 12:46:40 UTC