Forum

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 - mks_gis

Pages: [1] 2
1
TBH, I can't remember why it is False. This script develops as and when I need it, SfM is only a small part of what I do. I'll look into it when I next edit it, but you seem to have the skill to develop it further for your needs.

2
Thanks Paul, I'll look at your version and perhaps merge them.

The maintained version is on Github:
https://github.com/gisportsmouth/PhotoScan-Automation-Script


3
Hi Luiz,

 Thanks for your code. Not been able to do much on this lately, and to go through your code properly, but have updated to accommodate Metashape and some minor tweaks. Thought I'd share until I have more time.

Cheers
Martin


4
Hi Luiz,

 thanks for commenting. PhotoScan doesn't seem to update the points until the model is optimised.

 The approach in your second post works, but only until the session is shut down. Tracks would allow you to re-open the document later and carry on. But tracks give all points, what I'm really after is the valid points after alignment. I might not change to tracks because of this. I've made a few changes to the code and am testing, will possibly upload a newer version soon-ish.

Cheers

5
Currently needs to be run as one session, valid ties not accessible after
    gradual selection. To be fixed in next version using tracks.
   
    len(chunk.point_cloud.points) - valid ties
    len(chunk.point_cloud.tracks) - all ties

With gradual selection:
    After alignment the model is optimised and then a gradual selection is run.
    After the align process all tie points have errors attached
            to them. The values used are a matter of debate.
            Rec Uncert is normally set to 10 and run only once. If
            the value of 10 selects more than 50% of the tie points the value
            self-adjusts to less than 50%.
            Proj acc is set to start at 2 and self adjusts so less
            than 50% of the tie points left after Rec Uncert are selected.
            For Reprojection Error the aim is to select up to a set percentage
            (default 80) of the initial tie points. The value used is self
            selecting. The selection process is in steps of 10% and optimisation
            is run after each step.
            Values used are saved in the log file.
           
Without gradual selection:
    After alignment the model is optimised once.

3D Model/Geo: Geo basically is anything where the photos have coordinates or you are using Ground Control Points(GCP). If you have photos of objects without that info use 3D Model

Set-up workflow:
1. Set up your chunk(s)
    import photos
    name chunks (optional - helps in sorting the output)
    save a copy of the PS Doc

Custom workflow:
2. (Geo) Set reference system for each chunk
3. Load script
    chose output folder
    chose file prefix (this gets added to each output file - helps if trying different settings)
4. Under "Change Values" select "Get Current Parameter info" (check console window for output)
    check image quality threshold
    check filtering (options: Mild (default), Moderate, Aggressive, none)
    (Geo) check export CRS (in case you want the output to be different from the CRS of the camera/GCP)

No GCP or checking alignment - runs whole process and exports:
5a. (Geo) Choose "Custom":"Run - Geo" (with or without gradual selection)
or
5b. (3D Model) Choose "Custom":"Run - 3D Model" (with or without gradual selection)

With GCP or scales or if wanting to check alignment:
5. Choose "Custom":"Align only" (with or without gradual selection)
6. Set up GCP for each chunk or set scale
7. Check alignment
8a. (Geo) Choose "Custom":"Run all after alignment - Geo" (with or without gradual selection)
or
8b. (3D Model) Choose "Custom":"Run all after alignment - 3D Model" (with or without gradual selection)

Files generated depend on method used (3D or Geo) and input photos.

You can run methods individually as well after opening the script, in console type "ps_doc." and then hit "TAB" to see the options.

6
Python and Java API / Re: How do you access total tie point number in Python
« on: September 01, 2018, 04:06:06 PM »
Thanks Alexey!

7
fixed

needs to run as one session though, i.e if you've run the script keep the session open for grad sel process

8
General / Re: Your opinion on USGS Agisoft Processing Workflow
« on: August 31, 2018, 05:52:12 PM »
I understand commercial implications, it's a shame though as sharing our settings and experiences helps the community overall. We'll slog on, cheers.

9
Spotted a mistake, gradual selection currently only works for single chunks...

10
Python and Java API / How do you access total tie point number in Python
« on: August 31, 2018, 03:42:35 PM »
Hi,

 In the tie point information PS gives you y of x tie points. You can get the active number of tie points (y) by using len([i for i in chunk.point_cloud.points]). How do I access x, i.e the original number of tie points in Python, I can't find it in the reference doc?

Cheers
M.

11
General / Re: Your opinion on USGS Agisoft Processing Workflow
« on: August 31, 2018, 11:58:48 AM »

Funny, seeing how I followed this workflow and got worse results than following our own developed workflow.

I suppose it all depends on your data set.

Could you outline your workflow and describe your type of data, out of interest?

12
General / Re: Your opinion on USGS Agisoft Processing Workflow
« on: August 31, 2018, 11:08:11 AM »
Oh Geeze, the results after using USGS methodology are 100x better than not.
This should be part of the agisoft standard workflow IMO.

I wonder if the gradual selection steps could be automated. I am a .net guy mainly but can adapt to other languages.
I sent my "re-write" of the USGS workflow to agisoft sales. Hope they do something with it.
In the end, the PS results were better than Dronedeploy and pix4d.

I've automated grad sel here using Python, loosely based on this, but not fully (only do RecUncert/ProcAcc once each):

http://www.agisoft.com/forum/index.php?topic=9578.0

workflow with GCP:
run script
set export folder
set any file prefix for export files
in Custom Menu Align with grad
Do your GCP work
in Custom menu Run all after alignment - Geo with grad

Opinions welcome.

13
Python and Java API / tweaked automatic gradual selection process
« on: August 29, 2018, 01:16:34 PM »
***Please see later posts for updated script

I've tweaked the optional grad sel process:
1. rec uncert of 10, unless > 50% are selected, then self adjusting - optimise
2. proj acc self adjusting to select < 50% of ties left after 1 - optimise
optional break - allow for adding GCP
3. repro error self adjusting in 10% steps/optimise to select up to 80% of the initial ties.

Loosely based on USGS workflow discussed here: http://www.agisoft.com/forum/index.php?topic=9485.msg44134#msg44134

Seems to work with P4 images, Sony a6000 UVA images and DSLR close range models to reduce pixel errors. I know it's not a substitute for trying to tweak individual models for max accuracy, but good for quickly processing to a good standard, I believe. Happy to take comments.

Also added logging of values used in each step for reference if needed later.

14
General / Re: Your opinion on USGS Agisoft Processing Workflow
« on: August 24, 2018, 01:42:42 PM »
I'm unclear of one thing, amongst others. Perhaps someone can help.

In the gradual selection process they mention the percentage of points selected in each step. Is that percentage of total tie points you started with, or percentage of tie points left after each step. The latter obviously reduces the tie points considerably, when I tried it with some P4 data by about 90% of tie points we started with.

15
General / Re: Your opinion on USGS Agisoft Processing Workflow
« on: August 23, 2018, 05:37:16 PM »
This is really interesting! How would you update the workflow with the adaptive fitting option that PS now has? In the expanded document they say to use the default values or use the checked parameters for optimisation, which is best?

Cheers
Martin

Pages: [1] 2