Auto Link Control Workflow

[JamesHead-Mears1]

I'm in the process of evaluating ArcGIS Drone2Map as a potential replacement for our imagery processing software and was wondering if there is any additional information on using auto link control? There is some information at the following two pages:

• https://doc.arcgis.com/en/drone2map/latest/help/auto-link-control.htm 
• https://doc.arcgis.com/en/drone2map/latest/help/use-auto-link-control.htm

The second link has a paragraph at the end saying:

After the options have been configured, the next processing attempt will run the auto link control process and link any control points that it detects. Links can then be reviewed in the Image Links Editor window.

Does this mean after enabling the option it will auto link the imagery and ground control when I process the project?

I found that I could enable this option and once an alignment was completed, I could link a control point in the Image Links Editor and then it would let me Estimate and assign auto links for the other images. Is this the correct workflow or just another approach to using it?

[MarkBarker1]

Hi @JamesHead-Mears1,

Thanks for reaching out. Let me see if I can help clarify, there are 3 different ways you can do auto-linking in Drone2Map. Two of which involve deep learning and are in the topics you linked. 

- Assisted Links
- Auto Link
- Auto Link and Adjust

The first way (Assisted Links) does not require the deep learning installation, but is limited and requires more manual work. It is accomplished as such: 

1. With an open project, import your ground control.
2. Run the adjustment step (process with just the image collection enabled).
3. Open the Image Links Editor.
4. Place a link for the currently selected GCP. 
   

   Note: the Estimate Links button becomes available at the bottom of the window.

5. Click the Estimate Links button.
   
   Note: The image list will filter down to the image you linked and other images that have estimates. 
6. Click the  crosshair or the auto link button to accept an estimate as a new link.

The second method, Auto Link, actually needs to be enabled within the Options > Adjust Images > Auto Link Control section. This method uses deep learning models to review the images and detect the GCPs, cutting down on the amount of manual linking and reviewing you need to do. This option is good for when you want to review the automatically created links before they are incorporated into the adjustment, essentially quality control. The workflow would be as follows: 

1.  With an open project, import your ground control.
2. Enable the Auto Link Control option, set to "Auto Link".
3. Define any additional options.
4. Run the adjustment step (process with just the image collection enabled).
5. Open the Image Links Editor.
   Note: Images are already linked and ready for review
6. Once satisfied with the links, run the adjustment step again to verify accuracy.

The last option, Auto Link and Adjust, combines the previous workflow above with the adjustment step. Trusting the model to find and accurately link the points. You can still run a review of the links and readjust any that may have been detected poorly. However, you would need to run the adjustment step again if you modify any. This is why we recommend using the Auto Link option first as it is more efficient to not have to re-run the adjustment step.  

To answer your questions:

Is any additional information on using auto link control?
The topics you have found are the main information we currently have available. Although, we plan to update them again soon and potentially include the auto link workflow in a learn tutorial. 

Does this mean after enabling the option it will auto link the imagery and ground control when I process the project?
Yes, if you are using the Auto Link or Auto Link and Adjust options your ground control points will be automatically linked to your images. However, it is still highly recommended you review the links for accuracy before creating products. 

Is this the correct workflow or just another approach to using it?
The Estimate Links button workflow is different than the deep learning auto detection workflow, but either can be used to accomplish the same task. The ultimate goal being linking your ground control accurately. Regardless of chosen method, its important to review the processing report or Control Manager pane to assess accuracy values.

Appreciate the feedback on the documentation, I went ahead and updated it today with a bit of further clarification that will appear with the next release. 

Regards,
Mark

[JamesHead-Mears1]

Thanks @MarkBarker1. That's a great response!

I'm still not getting the Auto Links to generate after processing but I'm assuming that this is probably because the Cropped Image Size value is too low. Is there a way to show the position of the GCP in an image that hasn't been linked? This would help to figure out if the window being cropped is too small or something else is going wrong.

[MarkBarker1]

Hi @JamesHead-Mears1,

If you open the processing log while the initial processing step is running you should see indication of the detection process.

Its currently our main way to checking if the detections were successful. There are image chips created as part of the process but they are intermediate data so they get removed before finalization. It is a good idea though to keep them or have an option to do so. If you have a minute and wouldn't mind posting about it here: https://community.esri.com/t5/arcgis-drone2map-ideas/idb-p/arcgis-drone2map-ideas
We can maybe work it into our next release. 

Thanks,
Mark

[JamesHead-Mears1]

Thanks @MarkBarker1 . That helped me figure out the issue. I had to double the pixel size, and it started to detect them.