Units for gradual selection values

[aneverman]

Hi all,

Can anyone tell me what the units are for the values used in the gradual selection tool? As in, are these values errors in pixel distances, standard deviations, etc.

Thanks

[LFSantosgeo]

Hi all,

Can anyone tell me what the units are for the values used in the gradual selection tool? As in, are these values errors in pixel distances, standard deviations, etc.

Thanks

Have you found out?

[BobvdMeij]

There's been many forum posts on this subject in the past. Not surprising considering the relative vagueness on this matter provided by the User Manual. There's some interesting discussions going on in the below posts:

http://www.agisoft.com/forum/index.php?topic=6888.0
http://www.agisoft.com/forum/index.php?topic=5585.msg27338#msg27338
http://www.agisoft.com/forum/index.php?topic=3559.0
http://www.agisoft.com/forum/index.php?topic=6684.msg32291#msg32291

I must admit though that I too am still unsure what the best workflow is. Although I have performed countless iterations, each time altering different parameters, I find there's sometimes very little or no observable effect. And even if there is an affect it is unclear what is explaining these effects as certain parameters seem to work well for some products, but not so much for others.

[Smallpoly]

I find there's sometimes very little or no observable effect.

That's been my finding as well.. except when getting rid of low photo count images. Points with just two photos tend to be very unreliable, so I tend to remove them when I'll still have enough higher quality data left over.

[LFSantosgeo]

There's been many forum posts on this subject in the past. Not surprising considering the relative vagueness on this matter provided by the User Manual. There's some interesting discussions going on in the below posts:

http://www.agisoft.com/forum/index.php?topic=6888.0
http://www.agisoft.com/forum/index.php?topic=5585.msg27338#msg27338
http://www.agisoft.com/forum/index.php?topic=3559.0
http://www.agisoft.com/forum/index.php?topic=6684.msg32291#msg32291

I saw those posts and read carefully each one. But I'm unable to find any standard on how to operate the gradual selection parameters. Knowing the units would be a nice implementation for the next version of Photoscan.

I must admit though that I too am still unsure what the best workflow is. Although I have performed countless iterations, each time altering different parameters, I find there's sometimes very little or no observable effect. And even if there is an affect it is unclear what is explaining these effects as certain parameters seem to work well for some products, but not so much for others.

Could you share some of your findings in more detailed way?

I made some tests and without much effort I was able to improve my model accuracy even though not knowing how. The dense cloud seem to be affected in point numbers and total accuracy. As it is expected to as you eliminate the highest errors when filtering the sparse cloud.