星球

I am having issues with PixInsight+Seestar s50 mosiac (mainly stumped on ImageIntegration's error, but registration is being a pain). Any ideas?

Hello everyoneFor context, I began processing my images again and have learned PixInsight. My hardest project M 16 was completed in normal framing mode with 3.7k+ files with no problems. I then processed a couple other targets and just decided to tackle a ~15 hour 2x mosaic of the jellyfish nebula region, 12 hours LP (Ha-OIII dualband filter) and 3 hours IRCUT+ star spike mask for stars to emphasize the stars Tejat and Propus. The thing is, 2x mosiac for the seestar is not a 2 panel mosiac. My best description of it is a random density map (the seestar essentially moves around randomly in an enlarged field of view). Some frames are in the middle and some are barely even in this enlarged field of view, some may have like a third of the frame inside the 2x field of view. I call it a density map because the center mathematically should have more integration than a corner. After each session, I had save each frame on and it automatically outputs a master for that session. I definitely cannot use the IRCUT master as is because a plane went through that isn't super prominent but is still there, and it kept a couple pretty bad frames. The lp filter had a couple good masters and bad ones. Note I am confident that drizzle will be effective given my conditions and the ditheringNow for the problems and errors. Essentially, from the entire start of the preprocessing stage to the final part has some sort of an error with an exception to debayer and measurements at some point of my struggles. My usual preprocessing workflow is something like (in WBPP) debayer-> register -> local normalization -> measurements -> subframe selector -> image integration (pixel rejection) -> drizzle integration. Firstly, shoving in my frames in WBPP as they are will output a result cropped down to one frame. This means I have to use a manual reference frame. My first try in attempt 1 (on 3k frames for the lp filter) was debayer -> generate starfield in catalogstarfieldgenerator -> put it as a manual reference frame in WBPP and continue as a tutorial suggested. It failed with the ransac error unable to find and initial linear transformation or something like that. I got maybe 100/3k frames kept. I ran the same thing in staralignment with similar results. I then proceeded with the standalone process star alignment with the same reference frame with the same issue. I checked to make sure the psuedo reference frame had the right dimensions, rotation, size, and matched it to my stacked output from seestar live session. No matter what it seemed to fail and I gave up on it. I then used the stacked output as a reference frame and after multiple iterations it somehow worked. I used generate drizzle data in staralignment and unfortunately I do not remember the settings. I put the registered images from star alignment into WBPP for weighting and local normalization, and local normalization with some weird error but it turned out that I shouldn't have used it any way with a mosiac like mine so it didn't matter. I put it into subframe selector and the variance of certain statistics was extreme, but I figured out the outliers and rejected them after a while. Note at this point I have repeated all the somewhat successful tasks for both lp and IRCUT filters.Next was the part I can't get around: ImageIntegration. I have 16 GB of RAM and it worked perfectly fine for 3.7k files no mosaic with no issues. However, after inputting even the IRCUT data with only 1.2k files I got an insufficient memory error. I spent a long time trying to get around it. First idea was to close every little program and test various numbers working threads. No success. Then I inflated the backup disk RAM (or a "page file"). This is not risky to computer malfunctions and acts like a back up overflow RAM when it runs out. I played it safe and gave myself 16 gigs of back up, but PixInsight had a seizure when it got to backup RAM and the file explorer task ended and PixInsight crashed, leaving me to a grey screen. I have goofed up computers before, so I had a back up restore point before giving myself extra backup RAM and I was easily able to restart file explorer. I tried batches but nothing worked. I had the bright idea of skipping pixel rejection and putting the drizzle files into DrizzleIntegration, but it failed in 2.03 ms with some error that essentially meant either the drizzle files were all corrupted or missing infromation. At that point, I was back to square one. Attempt 2 has been much more of a pain in some ways. I updated WBPP so I can run subframeselector in WBPP and I have been working soley on the IRCUT data to find the best settings. No matter what I do, I can't even register the frames. My IRCUT is some of the best data I have worked with too, so I don't know whats up. I have run at least 20 different settings (haven't counted, but 3 days worth). Main questions: 1. How do I register my frames with this nonsense? 2. How do I get around the RAM issue, preferably without $?

讨论 1

喜欢