Tie points from the reference and target images #1
Assignees
Labels
enhancement
New feature or request
Comments
|
Hi Pedro, You're right, image-to-image registration literally is what you say, I use another "approach" is that base on the properties of the reference image the plugin adjusts the target image to align it, maybe I need to let it know this difference in the process in the readme file. However, this feature is really good to add to this plugin, any help is welcome. Regards! |
|
Thank you very much @XavierCLL ! This would be a great enhancement. Just for reference/orientation, some packages that do the co-registration:
|
|
Thanks so much Pedro for reference these two software to include as part of the new features, and to have a real co-registration plugin. I'm aware that this early version is the most basic approach and it's not literally image-image co-registration, I wanted to start with the basic, let's see how it goes. |
|
Hi @XavierCLL I have been working with AROSICS (https://git.gfz-potsdam.de/danschef/arosics) and I can tell that it performs a very good job co-registering several sensors, so it could be a good option. It requires several dependencies but using conda it's easy to install (https://danschef.git-pages.gfz-potsdam.de/arosics/doc/installation.html). Best regards! |
|
Good to know that you work for the AROSICS app, I took a look I found that is one of the best co-registration apps for remote sensing. I've been quite busy but this will be the first new feature for this plugin. Thanks! |
|
In the latest version (21.11.18), I've implemented the two algorithms of Arosics, global and local. I've tested it and works fine if you have the opportunity to make more tests, It would be great. Thanks. Regards! |
|
Hi @XavierCLL What a wonderful news! This is really awesome! I've tested here with QGIS 3.16.11, with Python 3.9.5 and it worked, both local and global co-registration! However, I could not run it with QGIS 3.22.0: Even more strange is that now I'm getting the same error with QGIS 3.16.11: I don't know if this is relevant, but I'm geting this Python warning when QGIS starts: Does these errors make sense to you? Thank you very much @XavierCLL !! |
|
Hi @XavierCLL and @PedroVenancio, pretty cool to see how this gets to life! I also had a short look but not much time for testing so far. However, as the developer of AROSICS, I would offer to test it and give you some feedback in the next time. I would really like to see this working! @PedroVenancio: The Python warning is not related to AROSICS - looks like QGIS is importing deprecated GDAL modules there. |
|
Hi @danschef I'm glad to see you looking at this! It's really great! It will open the use of AROSICS to a broad group of potential users! I think you are absolutely right, I was with py_tools_ds 0.19.0. Downgrading to 0.18.1, arosics global already works! Running local, I get this error, because it is expecting something that does not exist in version 0.18.1: I think the issue with gdal_polygonize can be related with the callback, as reported here: Meanwhile I've updated QGIS 3.22.0 to the latest version, and it seems that something had changed and some Python package is missing. GDAL, GEOS and PROJ had been updated, I don't know if it could be something related to that: |
|
Hi @PedroVenancio and @danschef nice to see you here and thanks both of you and your tests, any feedback is welcome to improve the plugin! @PedroVenancio open the Python Console in Qgis and run:
Copy the error here to check what is happening, thanks Xavier |
|
Hi @XavierCLL
Here is the output of QGIS 3.22.0: |
|
The issue with gdal_polygonize causing the PermissionError on Windows (#1 (comment)) should be fixed in py_tools_ds 0.19.1 which is already available on PyPI and will be released in the next hours on conda-forge. Regarding the installation of your environment: I can only recommend to use conda and install QGIS and AROSICS into the same environment: I just tested it with global and local co-registration and it works quite well. |
|
Thanks @danschef I've already tested py_tools_ds 0.19.1 with QGIS 3.16 and it works! However, I'm still getting this issue with arosics local but it works in another attempts. I didn't found the pattern yet. Meanwhile I tried it in conda QGIS 3.22 from OSGeo4W v2 is still failing. |
|
This AttributeError is caused by |
Looks like the answer is yes. I will see how I can fix this. |
I cannot reproduce that issue as well, but maybe because I've tested in an old VM Windows version. My primary OS, Linux, the plugin works like a charm (with parallelization for the local algorithm). I understand that the best option is using Conda to install Arosics and all of its depends, but most of the users (mostly windows users) is hard to say that to install a plugin it requires installing another software and running some commands in a terminal, and others steps. That is the reason that I'm trying to have a special zip with all dependencies inside for windows users to install it in the Qgis version installed in the system. I've updated the zip with the latest changes and with the latest version of Thanks guys! |
|
@PedroVenancio |
|
Hello team! Trying to use it for first time, using two PRISMA imagery i have the following. Any clue what is going on? |
|
Probably a memory issue. Could you have a look at your memory ressources when running this code? Warped PRISMA data can get quite large and it seems like you are using an in-memory output. Anyways, I don´t think this exception is related to the topic of the above issue. So you should open a separate one. I am also not sure if it is related to the code of the QGIS plugin itself or the underlying arosics package (which has a separate issue tracker). |
|
Hi @dpoursanidis, welcome! Thanks @danschef, yes it seems more about ram-memory or Arosics issue, but you can also test the plugin using the all-in-one zip to rule out a dependency issue: https://github.com/SMByC/Coregistration-Qgis-processing#windows |
|
Thanks @XavierCLL for the instructions. I have install it using the zip file. In the zip, at the help file there is a note "This plugin requires additional Python packages ( |
|
Weird, it should not happen, that is the idea with that package include all libs inside, are you using a 64bits Windows? |
|
I use 64bit windows, yes. |
|
Thanks for that. In that zip file, in the help, they have another link. Have you test it? |
|
You're right, that link expired, fixed, thanks! |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Dear @XavierCLL
Thank you very much for this plugin!
I was testing and I'm thinking if, besides the reprojection, resampling, extent adjustment and masking, it is not supposed to do the image-to-image registration, automatically taking tie points from the reference and target images and making the close alignment? This feature is not [yet!] implemented, am I right?
Thanks!
The text was updated successfully, but these errors were encountered: