Georeferencing and sharing historical air photos

385
9
Jump to solution
10-01-2019 12:34 PM
New Contributor III

I have a number of Historic Aerial images georeferenced that I'd like to clip, mosaic and publish for use in ArcGIS online either as a layer or a basemap using ArcGIS Pro 2.3.

The images have a collar and some collar information that can be totally removed. I have a fair amount of ArcGIS online experience but am new to desktop.

Can someone give some advice about how to go about doing this using ArcGIS Pro 2.3?

Are there any advantages to publishing as a basemap vs. a layer? 

Are there are ArcGIS Pro training plans that address this product/workflow specifically?

thank you

Reply
0 Kudos
1 Solution

Accepted Solutions
Esri Regular Contributor

Based on what you’ve said, I’ll give abbreviated advice

 

At a high level, there are three primary options, described here only briefly.  It sounds like you’re in the first case, and for the benefit of other readers who may see this, I'll describe the two other options afterward:

 

  1. Presuming a low number of photos and you do not need high accuracy, then use the Georeferencing tools.  Note this will not orthorectify the imagery (will not fit imagery to the terrain). Specifics:
    1. Georeference each image independently.
    2. Start with first order transformation to get started, but depending on your ability to find ad-hoc control points, use higher level transforms such as spline (a minimum of 10 tie points required to enable this transform). For reasonable accuracy you will want control points distributed around image and on terrain extremes.
    3. Use “Save” which will georeference using tie points in a small *.aux.xml file that can be edited (improved) later if necessary.  In this scenario recommend you do not use “Save As” which creates a new image file.
    4. Create a mosaic dataset to dynamically mosaic the images together.  (any image can be viewed, so pixels in overlap are not lost).
    5. Use Footprint feature class to clip the black collar (pixels not permanently removed, just hidden)

       

  2. For highest accuracy, scalable to large blocks of film, users should apply photogrammetry via the Ortho Mapping capabilities (ArcGIS Pro Advanced license, Scanned Film workflow)
    1. Should have the camera report to define parameters such as focal length - this will most likely be available from USGS Earth Explorer
    2. Need to create* a table with approximate exterior orientation listing the (x,y,z) location and camera heading angles (kappa) for each image. This can be estimated if necessary.  May be created manually for small to medium size blocks of film.
    3. Need to provide a DEM – this can be exported from ArcGIS Online World Terrain image service, use an existing DEM, or generate a DEM from the imagery after block adjustment
  3. If you have limitations on inputs for #2 for Ortho Mapping (e.g. no exterior orientation table and/or not able to see fiducial marks on film) use the Drone workflow in Ortho Mapping to generate an approximate mosaic.
    1. For some use cases, this may produce adequate products.
    2. For users seeking the highest accuracy results, use results of this process as input to the Scanned Film workflow in Ortho Mapping

 

We’ll be publishing a deeper discussion on this on http://esriurl.com/ImageryWorkflows soon - I will plan to update this post when that is published. 


Cody B

View solution in original post

9 Replies
Esri Regular Contributor

Ryan

(I took the liberty of editing your title so that others may find this post - I hope you don't object)

Short answer, yes.  

Unfortunately I'm rushed for time today, so I may not be able to answer more fully until tomorrow, but yes you can do this and we can provide a workflow.  

My recommendations will depend on how many photos you have, and what accuracy you seek to achieve.  If your screenshot shows the entire collection and you do NOT expect high accuracy, the easiest approach is to use the Georeferencing tools.  (see Help)

If you have more photos and/or you're seeking high accuracy, you'll use Ortho Mapping (ArcGIS Pro Advanced license required, Pro 2.4 recommended).  In that case, the black collars will be critically important, to enable you to find the fiducial marks around the perimeter of the image.

Check back here and I'll provide more information soon

Cody B. 

Reply
0 Kudos
New Contributor III

Thanks Cody, I don't mind the title change.  I have 38 photos, about 100mb each from USGS Earth Explorer and I've already georeferenced them individually using the georefernce tools.

The primary purpose of using the imagery is to quickly determine if a road was in place at the date the photos were flown.  They are also useful for finding re-alignments of the roads so being able to turn them off and on and compare to existing roads today is helpful. 

With that in mind I'm guessing roughly +/- 50' or so seems like it would be accurate enough for the intended purposes without being too disorienting to the user when flickering off/on or dimming with transparency and comparing to the roads today. 

I'm attaching a screenshot of one of the full images in case it's helpful.  Because I have significant overlap in the images and they are all flown on the same day, I don't believe I need any of the collar information including what you see on the right side of the image below which is written on top of the image rather than in the collar.  I believe the neighboring images will allow that portion to be clipped out without any loss of imagery.  thank you

Reply
0 Kudos
Esri Regular Contributor

Based on what you’ve said, I’ll give abbreviated advice

 

At a high level, there are three primary options, described here only briefly.  It sounds like you’re in the first case, and for the benefit of other readers who may see this, I'll describe the two other options afterward:

 

  1. Presuming a low number of photos and you do not need high accuracy, then use the Georeferencing tools.  Note this will not orthorectify the imagery (will not fit imagery to the terrain). Specifics:
    1. Georeference each image independently.
    2. Start with first order transformation to get started, but depending on your ability to find ad-hoc control points, use higher level transforms such as spline (a minimum of 10 tie points required to enable this transform). For reasonable accuracy you will want control points distributed around image and on terrain extremes.
    3. Use “Save” which will georeference using tie points in a small *.aux.xml file that can be edited (improved) later if necessary.  In this scenario recommend you do not use “Save As” which creates a new image file.
    4. Create a mosaic dataset to dynamically mosaic the images together.  (any image can be viewed, so pixels in overlap are not lost).
    5. Use Footprint feature class to clip the black collar (pixels not permanently removed, just hidden)

       

  2. For highest accuracy, scalable to large blocks of film, users should apply photogrammetry via the Ortho Mapping capabilities (ArcGIS Pro Advanced license, Scanned Film workflow)
    1. Should have the camera report to define parameters such as focal length - this will most likely be available from USGS Earth Explorer
    2. Need to create* a table with approximate exterior orientation listing the (x,y,z) location and camera heading angles (kappa) for each image. This can be estimated if necessary.  May be created manually for small to medium size blocks of film.
    3. Need to provide a DEM – this can be exported from ArcGIS Online World Terrain image service, use an existing DEM, or generate a DEM from the imagery after block adjustment
  3. If you have limitations on inputs for #2 for Ortho Mapping (e.g. no exterior orientation table and/or not able to see fiducial marks on film) use the Drone workflow in Ortho Mapping to generate an approximate mosaic.
    1. For some use cases, this may produce adequate products.
    2. For users seeking the highest accuracy results, use results of this process as input to the Scanned Film workflow in Ortho Mapping

 

We’ll be publishing a deeper discussion on this on http://esriurl.com/ImageryWorkflows soon - I will plan to update this post when that is published. 


Cody B

View solution in original post

New Contributor III

Thank you Cody,  I think the first workflow will work for what I'm trying to accomplish.  I was able to create the mosaic but I got lost on part 5, "Use Footprint feature Class to clip the black collar"  Is that a function in 2.4?   I don't have footprints so I suppose I'll need to create one based on the extent of the raster and then modify it (shrink it) to hide the border?

I can wait for further explanation in the published workflow if you are going to go into further detail on those steps.  thanks

Reply
0 Kudos
Esri Regular Contributor

Ryan

I'm guessing you're not familiar with the Mosaic Dataset - please see link above.  (I was not referring to the GP tool "Mosaic To New Raster" - that creates a new raster file where any pixels hidden in the overlap are gone forever)

The Mosaic Dataset is a virtual mosaic, managed in a geodatabase (the images are not in the GDB, just the MD), and the Footprint is a built-in component that defines the extents of each individual raster. 

Look for "Build Footprints", use Method = "None" and use the Shrink Distance parameter to reduce the size of the Footprints (units are meters).  Then make sure the MD property under Defaults for "Clip to footprint?" is enabled.

Note you can edit them manually but since you have so much overlap you should be able to shrink them by ~500 to 1000 meters (a guess)

Cody

New Contributor III

Thank you that worked great!

Reply
0 Kudos
New Contributor III

Cody, any chance you could elaborate on how to go from the Mosaic to ArcGIS online while minimizing credit consumption?  I've got about 30 images totalling around 5 gb georeferenced and mosaiced following the steps we discussed above.

Reply
0 Kudos
Esri Regular Contributor

Ryan

sorry I couldn't reply sooner; this will be very brief

I'm pretty sure this is correct but I don't have time to test right now.   Could you try it and let me know?

  1. Manage Tile Cache—Data Management toolbox | ArcGIS Desktop  to create a tile cache on your computer
  2. Export Tile Cache—Data Management toolbox | ArcGIS Desktop  to create a "package" (*.tpk) from the cache
  3. Share Package—Data Management toolbox | ArcGIS Desktop  to upload and publish the TPK

(I don't recall if step 3 actually completes the publishing process, or if you have to go to a 4th step, e.g.

   4. log into ArcGIS Online, locate the package, and publish...)

This WILL charge you credits for data storage on ArcGIS Online; 1.2 credits/GB/month = $0.12.  Presuming your cache is approx. 5 GB this will be $0.60/month - cost doesn't change if 1 million people view it, or nobody.  Your concerns over credit usage may be a reference to using ArcGIS Online to create the cache; that does cost credits and surprised some users by consuming more credits than expected - but once ArcGIS Pro (and ArcMap) had the ability to create cache on the desktop (several years ago), in most cases it's less expensive to perform the caching process on your desktop, then upload & publish the cache.

Let me know your results

Cody B

New Contributor III

Thanks, i appreciate the response.  I'll try it out hopefully this week and report back.  The info on credits is really helpful.

Reply
0 Kudos