Messages

1

General / Re: Any way to disable internal compression entirely?

« on: July 12, 2018, 08:10:27 PM »

Appreciate the quick responses!

The ortho being shown in the last screen shot is the result of a raster transform between the dark channel and the brighter channel (NDVI specifically).

I think it is showing up only as a result of the NDVI formula which is making subtle changes between the two channels much more apparent.

Would the Assign Image option apply this same transform to the images being previewed?

2

General / Re: Any way to disable internal compression entirely?

« on: July 12, 2018, 05:25:01 PM »

I thought it might be a seam line as well, but there is only one seam line that runs vertical through that polygon. The blending mode is mosaic; would that be doing blending well away from the seam line?

Unfortunately I had to quickly redo this project a different way, so it would take me a bit to recreate to show the seam lines.

This was showing up because one of the multispectral channels had a very low signal (pixel values of ~20-30 of 255), and the other was very high.  The low signal channel was set as the primary channel had its brightness settings turned up to 400% in order to process. This showed up on 3 different processing runs of different areas, but using data from the same flight where one of the channels had low pixel values.

I did a test where we brought that specific channel up (histogram expanded it) on the images in the area, and the issue has gone away. That sort of correction however can't be used for NDVI though, as it would produce non-radiometric changes. Although the channel is dark, the signal is well above the noise.

Unless mosaic blending is creating a blocky area like that from two images (seems unlikely), I'm fairly positive we're looking at jpeg artifacts.

3

General / Re: Any way to disable internal compression entirely?

« on: July 12, 2018, 12:47:38 AM »

Here is a more apparent example. This is a gray NDVI image as shown inside Photoscan. The top image shows what happens when you select a replacement image over the ortho. The underlying frames do not have the artifacts, but the ortho does.

4

General / Re: Struggling to scan small plastic knob

« on: July 11, 2018, 09:41:09 PM »

That's not half bad actually. Dense 3d reconstruction is matching pixels from one view with pixels from another view. If the areas are smooth, low texture, then 3d information is difficult to reconstruct. The inside of the grooves looks pretty uniform. Shiny also makes it worse.

You might get a better model if you can find a way to coat the knob in something that produces visible patterns on the low texture areas. Sometimes spray paint can help here to reduce gloss. Might also try taking some sand paper to the knob and lighty sanding everything, including the inside of the grooves, to give it texture. Not enough to change the shape.

Otherwise, a different 3d reconstruction technique would work better (project light, line laser).

5

General / Any way to disable internal compression entirely?

« on: July 11, 2018, 05:47:24 PM »

I have been able to identify artifacts in the final ortho products that are due to the use of jpeg compression internally on the generated mosaics. The input frame data is losslessly compressed, and the exported tif files have lossless compression as well, but the internal tif tiles (by looking through the Photoscan project diretory) use jpeg compression.

In 1.3.5, the Project Compression slider can be set to zero, but it doesn't remove the jpeg compression artifacts. In 1.4.x, the setting is entirely gone.

Is there any way to force Photoscan to not use jpeg compression internally?

This significantly negatively impacts the accuracy of resulting NDVI images. The jpeg blocking is very apparent.

6

Python and Java API / Re: Dense cloud not in project after processing

« on: June 28, 2018, 09:57:13 PM »

The issue in my case was the frequent use of save( path ) for an already open project. Thanks to Agisoft support, it was identified this was causing the chunk objects to be invalidated. Only save() should be used if there is an open project and the path isn't being changed.

7

Python and Java API / Re: Dense cloud not in project after processing

« on: June 28, 2018, 04:32:03 AM »

I'm seeing a similar behavior on versions 1.3 and 1.4. I have a script that does the alignment step (matchPhotos then alignCameras) and it clearly spends a lot of time processing and the logs show it works, but when it finishes, there are no tie points and the cameras don't look aligned.

If you do the same steps over again but with the UI directly, it works.

This has been observed in both 1.3 and 1.4, and I have a setup where I switch between them occasionally (primarily for testing and development).

8

General / Re: Force selection of pairs to solve?

« on: June 28, 2018, 01:09:14 AM »

I believed I tried this last year and there was indication it was still using pairs that I didn't specify (as seen by view matches). I will try again with the latest version and make sure to explicitly disable the other preselection types.

This is 6 band data (Color, NIR, Red edge, thermal) from a dual swath camera configuration, so 8 cameras. All images are precisely exposed simultaneously.

9

General / Re: How to use pre-aligned multiband images?

« on: June 28, 2018, 12:38:54 AM »

Upon further testing, it looks like the tif files need to be internally tiled, not stripped, to correctly load into Photoscan.

10

General / Re: How to use pre-aligned multiband images?

« on: June 27, 2018, 05:47:39 PM »

Alexey, appreciate the response.

I'll give this a try again. I had a multichannel tif image setup earlier (a geotiff actually), but Photoscan couldn't view the image (it would load into the program fine).

Are there any requirements for the internal tiff structure (blocks vs scanlines, pixel interleaving?

11

General / Re: Force selection of pairs to solve?

« on: June 27, 2018, 03:24:30 AM »

This is why this is needed. It solved the cameras, but all the time coincident images generated a ton of bad data and definitely impacted the final geometry solution.

12

General / Force selection of pairs to solve?

« on: June 26, 2018, 11:32:58 PM »

If I have the ability to select pairs I want to Photoscan to use, can I force it to only use those pairs for feature detection, matching, and camera alignment?

I can verify that the pairs being selected would provide sufficient connectivity. It would be important that only those pairs get used, and no others.

This is because Photoscan fails to align datasets that frequently have either time or location coincident images where the pair matches do not have good 3d information (no baseline).

Alternatively, is there a way to force Photoscan to not attempt to align time or location coincident images? I understand it probably doesn't have image acqusiition time, but a simple way to do this for location would be to not use any pairs that are not sufficient far enough away from each other.

13

General / How to use pre-aligned multiband images?

« on: June 26, 2018, 11:29:46 PM »

I'm attempting to process a multispectral dataset using a set of tif images that can either be multi-page or multi-channel (It appears only multi-page is supported?).

Each page/channel is previously band aligned, so only the first/master channel would need to be used.

However, it appears in any multispectral chunk setup, regardless of multiplane or multilayer, it appears to be detecting features and attempting to realign all the other channels.

Is there anyway to force Photoscan in the scripting interface to stop doing this, and only use the primary channel for camera alignment?