Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Messages - stihl

Pages: [1] 2 3 ... 28
General / Re: Missing elevation points in DEM
« on: August 28, 2020, 10:44:54 AM »
This specific issue is separate from the tie-points or the alignment. The features (i.e. the branches and leaves) of these specific plants are too difficult for the dense cloud algorithm to match.
What you can try is reposition the bounding box so that it envelops the region where the vegetation is problematic and re-run that specific region with a lower quality dense cloud. Reducing the quality level effectively means that Metashape down-samples the original images before running the dense cloud algorithm. Because the resolution is lowered, small features such as leaves and branches might come out better in the dense cloud because small (pixel level) deviation between the leaves and branches in consecutive images becomes smaller and thus increases the chances of successfully reconstructing the plant, albeit in a lower level of detail, but a lower level of detail is always better than no detail. :-)

The approach you should follow in order to reduce the deviation on the GCPs is by filtering the tie-points ór heavily reducing the 'weight' (standard deviation) of the camera positions. Since you're using a P4 RTK it's best to use the gradual selection tool to reduce the amount of (bad) tie-points.

Filtering the tie-points in essence also reduces the weight of the camera positions as less measurements (in the form of tie-points) are available, which increases the importance of the GCPs.

Use the gradual selection tool to filter tie-points visible on only 2 images with the 'Image count' tool, then use the other tools; 'reprojection error', 'reconstruction uncertainty' and 'projection accuracy' to reduce the bad tie-points further. Right click on your chunk in the workspace pane and select 'Show Info' to check the RMSE of the tie-points. Preferably get this RMSE value below 0.8 pixels.

General / Re: DJI P4 RTK coordinate discrepancy
« on: August 28, 2020, 10:29:07 AM »
HI someassiedude,

How long did you let your DJI RTK base log while flying your project? Was it one hour, two hours or more?
Do all the flight positions show the same kind of deviation (mostly in the Y-axis/Northing) when you fix your model to your checkpoints (as a test)? Or instead - do all the checkpoints show the same kind of deviation?

General / Re: Delete or Moved a or few markers AFTER you orthomosaic ?
« on: August 27, 2020, 01:08:42 PM »
As markers (GCPs) affect the tie-points and internal and external parameters of the images, which in turn affects the dense cloud, which affects the mesh and thus the orthomosaic - it is necessary to go back to the Build Dense cloud after adjusting the markers. It is also necessary to either update or optimize your project after making any adjustments to the GCPs.

General / Re: Missing elevation points in DEM
« on: August 27, 2020, 01:04:04 PM »
It appears that the dense cloud algorithm was unable to find any features on those specific plants. Perhaps due to wind that the branches and leaves moved too much in overlapping consecutive images. Does this issue appear in other regions of your model as well?

General / Re: Best Practice for DEM "Resampling"
« on: August 27, 2020, 01:01:24 PM »
Hi to both.

The DEM resolution is inherently and directly tied to the Dense Cloud point density. For example - if your project was flown on 1 cm GSD and you process your dense cloud as Ultra High quality, this means that the resulting dense cloud will have an average point density (sampling resolution) of 1 cm (i.e. the points are on average 1 cm spaced from each other).

If you process your project on High quality (instead of Ultra High) then this point density goes down by a factor of 2. If you process it on Medium quality, the density goes down by a factor 4. Low is a factor 8 and Lowest is a factor 16.

The Build DEM step generates the highest possible sampled DEM based on the point density of the Dense point cloud. Meaning that if your project has a 1 cm GSD - if you process your project on Ultra High, the resulting DEM will have a default sampling resolution of 1 cm. High produces a DEM resolution of 2 cm, medium produces it on 4 cm, etc...

The next step is exporting the DEM from Metashape in any arbitrary resolution that you want. Such as 1cm, 2cm, 3cm, 4cm .. as long as it's a lower resolution than the DEM produced by Metshape. So if you have a 1 cm GSD project and wish to generate a 3 cm DEM - simply process a High quality dense cloud, generate a 2 cm DEM (default values) and export it as a 3 cm DEM geoTIFF.

General / Re: corn field impossible alignment
« on: August 27, 2020, 12:51:53 PM »
Hi Tuffi,

What is the resolution (GSD) of the images? If they're taken at a low altitude and every image has a lot of detail of the individual plants then this can have the adverse effect of each corn crop being in a slightly different location (view) due to factors like wind. This will mean that Metashape has a lot of issues in finding matchable tie-points due to the images not looking similar enough.
You can try rerunning the alignment on a lower quality level. This means that MS will downsample the images before finding tie-points. If the resolution is lower, then small differences in consecutive images will be reduced which can yield a better alignment.

If you've flown at a high altitude but still get a bad alignment then it's possible that MS will not be able to find proper tie-points and perhaps then it's necessary to refly the project with a higher overlap or by adding oblique imagery.

General / Re: Aligning Aerial Photos of Forest Canopy
« on: October 25, 2019, 05:06:58 PM »
Hi James,
I've always gotten much better results with dense vegetated areas when using downsampled images. This is due to there being less pixel variance on lower resolutions on vegetation which causes Photoscan to find more reliable tie-points and thus is able to align the images. This also accounts for the densified point cloud.
If you change the accuracy parameter (from High to Medium or even Low) then Photoscan will first downsample the input images before finding tie-points.
In the case that this does not give suficient results then it's also worth while to manually reduce the size further in an external application and then using these lower res images again (with high, medium and or low parameter tries).

I've gotten very good results from a nearly useless data set in Nigeria by downsampling the input images in Photoshop before bringing them back in to Photoscan and running it on Low quality dense cloud.

General / Re: Effects of the quality of dense cloud generation
« on: June 19, 2019, 05:42:42 PM »
Changing the quality level from High to Medium to Low changes the resolution of the input images.
If you select Medium, Photoscan will first downscale the images to 1/4th of their original resolution. 1/4th of the original pixels available results in a dense point cloud with 1/4th of the amount of points when compared to processing the project on High.

Per Image Side Downsample Factor                               Per Image scaling            Expected point density
Ultra High Quality:                          1:1                                      1:1                                             GSD        1:1
High Quality                                        1:2                                      1:4                                             GSD        1:2
Medium Quality                                1:4                                      1:16                                          GSD        1:4
Low Quality                                         1:8                                      1:64                                          GSD        1:8
Lowest Quality                                  1:16                                   1:256                                       GSD        1:16

Example: If Ultra high produces 100 million points, Medium quality will then create 25 million points.
If your project GSD is 2 cm/pixel, your maximum DEM resolution at Ultra high will be 2 cm/pixel. At Medium it will be 8 cm/pixel.

If Photoscan has to work with lower resolution images to create the dense cloud (and geometry of objects), then this will have an impact on the quality of the reconstructed geometry. It's best to make a small bounding box in your project area and test different quality levels to see if the low quality parameter provides you with the level of detail you're looking for.

- accuracy of 0.001 mm doesn't seem to reasonable value selected, as the nodes of the printed checkerboard pattern cannot be measured with such precision (I would suggest to use 0.1 mm max),
Can you please explain what changing this value in metashape actually has an impact on in the generation of sparse / dense cloud process? Does this affect the SIFT algorithm somehow or the bundle adjustment? How does having a grosely wrong accuracy value come into play in the reconstrution?
The accuracy of the markers directly impacts the 'weight' of the manually placed markers. If a manual marker is slightly off from the actual position then this introduces a misalignment of the images due to the weight of the markers being higher than the weight of the tie-points found by Photoscan.
The alignment determines the location and orientation of each aligned image. When building a dense cloud, an algorithm is used to create depth maps from stereo paired images. If the images are slightly misaligned then this will introduce noise in the final point cloud.

You will need to use the optimization tool to refine the camera calibration parameters. As currently there are significant lens distortions which affect your products.

Hi szhu0,

If you already have a DEM TIFF, why do you want to merge it with the XYZ data? From what I understood these two files are of the same project/data set? If so then there's no need to use the XYZ. The DEM (elevation) geoTIFF file will have the same kind of information, if it's in the same resolution.

You can assign a hillshade and colour ramp to your DEM TIFF in ArcGIS by doing this:

General / Re: Dense Cloud Quality
« on: February 28, 2019, 06:43:48 PM »
 -Photo Resolution Scaling:  Alignment Quality-
Highest:        2:1
High:               1:1
Medium:       1:2
Low:                1:4
Lowest:         1:8

-Photo Resolution Scaling:  Dense Cloud Quality-
Ultra High:      1:1
High:                   1:2
Medium:           1:4
Low:                    1:8
Lowest:             1:16

Example: If Ultra high produces 100 million points, Medium quality will then create 25 million points.
If your project GSD is 2 cm/pixel, your maximum DEM resolution at Ultra high will be 2 cm/pixel. At Medium it will be 8 cm/pixel.

General / Re: inaccurate volume
« on: November 12, 2018, 06:40:15 PM »
If you have not scaled your project with either markers (GCPS) or a sale bar then your model will have *no* value other than looking pretty.

Photoscan has no idea what the size of the box is unless it's scaled. Otherwise it can be a box the size of ..a box, or a box the size of a house.

Pages: [1] 2 3 ... 28