Question about optimizing alignment

[andyroo]

I was wondering if multiple optimization steps while entering GCPs is a bad idea. With a ~30 km-long strip, I find that the "smile" artifact with initial alignment makes GCPs get progressively farther off from their projected locations when I move from one end of the dataset to the other after I initially update the georeferencing by defining a few points as I discuss here.

So far I have resisted the optimization step until after I register all GCPs, since I figured it would degrade alignment if I do the optimization step multiple times. Not knowing the intricacies of bundle adjustment mathematics, I would love some insight into this. If I could do one optimization after entering a few GCPs along the model's length, then another when complete (possibly a third) this would speed up the GCP alignment by a lot (I have about 200 GCPs and prefer to manually align them on each image, resulting in generally about 1000 GCP alignments steps on ~1000 images).

[Porly]

Thats a good question.

In theory it could work, since BBA minimizes the error under weights of all GCP in each optimizing step. On the other hand, the quality of approximated model coordinates (GCP) would be different what could cause maybe a worse result. Few points  of good quality after first optimization, the rest better quality but still poor.

I tried to figure it out in the PhotoScan version before 0.9.1. Optimize with GCP's in border regions to get better location quality for all GCP's in images. Then i marked the rest and optimized it again. The result was different compared to marking all GCP's after alignment and then optimize all togehter.

You can try it and compare the results. But maybe it could work now in this version.

BBA under weights of GCP before version 0.9.1 was a lot of "weaker". The influence of the GCP's is higher now. Even with poor coordinate quality you can reach a deviation below 5CM (what I find a little bit unreal).

Before version 0.9.1 everything below 10 - 15 CM (after many many optimizing steps /w cleaning sparse point cloud) was a real good result in my photosets. Note that my normal sets have +- 1-2 square km. The error propagation in your case (30km strip) should be higher.

 

[Porly]

One thing that could help you:

If you have an old project of the same or similar scene (30km strip), you can use the camera calibration file of this project (after final optimization) and use this as intial parameters (or even fix) before alignment of the new photoset. That will reduce the bowl effect a lot what causes a much better result (when same camera was used). Later you can optimize with checked parameters (cx,cy,fx,fy,k1,and so on).

I have tested it. Worked fine for me.

[andyroo]

Thanks for that camera calibration parameter tip! I'll try that out this evening.

[Porly]

ok, give me a feedback if it helped or not

[andyroo]

Hi Porly,

Thanks a bunch for that tip about taking calibration parameters from a previous flight. the bowl effect is removed and GCP placement is much easier.

[Porly]

no problem, i am glad that it worked in your case too!