Align photos with local coordinate is not good

[ugo]

Hey guys,
i want use the markers (no coordinates) to align the photos. But that's not worked.
Did someone know what cani do for it?

thanks

[Alexey Pasumansky]

Hello ugo,

Are you using masks for this project?

[ugo]

Hello ugo,

Are you using masks for this project?

Yes, i used it.

[Alexey Pasumansky]

Hello ugo,

Maybe you can send the project file (re-saved as PSZ) to support@agisoft.com? If possible to provide also the original images (or converted to the JPGs), it would be also helpful.

[James]

My guess is that you have way more false positive matches than actual positive point matches.

The repeating pattern (including the human like figures) is probably matched with itself at multiple places and due to the small number of photos and lack of texture elsewhere on the item there aren't enough true matches to detect which the false matches are.

I suggest the solutions would be one of the following (probably do both 1 and 2):

1. reshoot with more like 30 photos around the circumference so that actual corresponding points are more likely to be matched due to the smaller angular interval so matching features look more similar between consecutive images.
2. reshoot with the item placed on a textured base like a piece of wood or newspaper (which is rotated with the item) to increase the chance of making real positive matches between points
3. keep on adding markers until there are enough to outnumber all the false matches (this will take a long time!)

There is a way to see which images/points are matched (right click on an image in the photos pane and select 'View Matches'), then you will see that points are matched between images that do not overlap (if i'm right).

[ugo]

Hello ugo,

Maybe you can send the project file (re-saved as PSZ) to support@agisoft.com? If possible to provide also the original images (or converted to the JPGs), it would be also helpful.

I think i find the reason. The Photos same positions from different perspective have the different color (different reflex color).

[ugo]

My guess is that you have way more false positive matches than actual positive point matches.

The repeating pattern (including the human like figures) is probably matched with itself at multiple places and due to the small number of photos and lack of texture elsewhere on the item there aren't enough true matches to detect which the false matches are.

I suggest the solutions would be one of the following (probably do both 1 and 2):

1. reshoot with more like 30 photos around the circumference so that actual corresponding points are more likely to be matched due to the smaller angular interval so matching features look more similar between consecutive images.
2. reshoot with the item placed on a textured base like a piece of wood or newspaper (which is rotated with the item) to increase the chance of making real positive matches between points
3. keep on adding markers until there are enough to outnumber all the false matches (this will take a long time!)

There is a way to see which images/points are matched (right click on an image in the photos pane and select 'View Matches'), then you will see that points are matched between images that do not overlap (if i'm right).

Thank you so much.