solving a sequence of photos

[Per Ljung]

Hi Dustin - 

I have a noob question on solving a sequence of photos. Using a static wide-angle camera I take a sequence of 1000 photos with 1sec exposures. Using image magick I can create a nice looking composite star streak image. Using celestial navigation algorithms I can compute the camera position for each photo after identifying the stars. My problem is that the solve-field results are rather noisy.

When I (locally) independently solve for each photo (typically >100 stars) this results in separate SIP2 coefficients for each photo, where the SIP coefficients are similar but noisy. The identified photo center is also noisy. Plotting the corr files field values results in noisy ~3 pixel wide star streaks. The corr index star streaks are even noisier. My CelNav algorithms solve for the lat,lon of each solved photo, but the noisy solve results in undesired noise in the lat,lon solution. 

To get less noise, can I re-use the SIP coefficients from the 1st photo for all subsequent photos?  If so then how? Or is there a better way to analyze a sequence of related photos from a static camera?

Here are some images illustrating the noise. 

https://drive.google.com/drive/folders/1aG7vKrC0O9x3nkXfjMQ4ZrdRr1jZKfvs?usp=sharing

streak.png

streak_field_index.png

field_index_detail.png

Thx