Skip to main content
Solved

Affin or offset polygons to the pixel coordinates of the corresponding aerial image

a_afiaax
Contributor
Forum|alt.badge.img+2

Hello FME community,

 

I am currently working on a deep learning project that involves preparing training samples for building polygons. 

 

I have the building polygons in an OGC GeoPackage file with Coord.System EPSG: 26916, as shown in following screenshot: 

image_001 

Additionally, I have the PNG aerial images for the same area without projection, which means every aerial image X,Y starts from the upper left (0,-0) and ends at the lower right (2000,-2000) as shown in following screenshot:

image_002 

I need to have the polygons of the training sample in pixel coordinates for each corresponding image. The output should be a CSV file that includes the vertices of every polygon in pixel coordinates as shown in following screenshot: 

image_003 

I have attached the datasets for better understanding.

Does anyone have any ideas or potential solutions?

Best answer by daveatsafe

a_afiaax wrote:

Hello Daveatsafe,

Thank you for your helpful solution and the attached FME-workbench.

However, I encountered an issue where the offset-transformer moved all polygons to (0,0) of the projection of all images not to the (0,-0) of every non-projected PNG image, and they are overlapping each other , as shown in this screenshot:

image_00 

Each image is 2000 pixels * -2000 pixels, and each polygon corresponds to a specific house on an image, as shown in this screenshot:

image_01 

Before extracting the pixel coordinates of every polygon, I need to offset the projected polygons of every image to the right place in the their associated non-projected PNG image (2000 pix * -2000 pix and 0,-0 origin for every PNG), as shown in second screenshot.

 

Do you have any suggestions on how to achieve this?

I have attached the Waukesha projected TIFF images i used, corresponding PNG images for deeplearing and the FME-workbench for better understanding.

Having the GeoTIFFS is a great help, since they contain the georeferencing needed to align to the polygons. I am attaching a workspace that will extract the origin and pixel size from the GeoTIFFs, then use this info to move and scale the polygons to align to the pixel space coordinate.

View original
1 Attachments
Did this help you find an answer to your question?

4 replies

daveatsafe
Safer
Forum|alt.badge.img+19
  • daveatsafeSafer
  • Safer
  • June 26, 2023

Hi @a_afiaax​,

The key step would be to determine the georeferencing of the images.

Based on the image names, they are part of an orthophoto tile set, but unlike most tiling schemes there appears to be some overlap between images 1_34_21 (south) and 1_34_20 (north). Being PNG files, they would have no internal georeferencing, and would need sidecar PGW or WLD files. Without this georeferencing, there is no way to tie the building outlines to the imagery.

I took a look at the Waukesha County GIS Portal, and found that their interactive orthophoto download only offered PNGs without any georeferencing, which isn't that helpful. However, they did offer MrSID tiles through their Elevation and Imagery download page (https://gis2.waukcogeo.com/portal/apps/webappviewer/index.html?id=32f4e52f72d642deb67d0f330db1ab1b). MrSID files are FME-readable, and are properly georeferenced.

I am attaching a workspace to extract both PNG coverage of each building from the MrSID file and the Excel WKT for the polygons within that image. Some of the buildings lie partially outside the ortho extents, and may give incorrect images. The WKT coordinates will be from the image lower left.

a_afiaax
Contributor
Forum|alt.badge.img+2
  • a_afiaaxContributor
  • Author
  • Contributor
  • June 28, 2023

Hello Daveatsafe,

Thank you for your helpful solution and the attached FME-workbench.

However, I encountered an issue where the offset-transformer moved all polygons to (0,0) of the projection of all images not to the (0,-0) of every non-projected PNG image, and they are overlapping each other , as shown in this screenshot:

image_00 

Each image is 2000 pixels * -2000 pixels, and each polygon corresponds to a specific house on an image, as shown in this screenshot:

image_01 

Before extracting the pixel coordinates of every polygon, I need to offset the projected polygons of every image to the right place in the their associated non-projected PNG image (2000 pix * -2000 pix and 0,-0 origin for every PNG), as shown in second screenshot.

 

Do you have any suggestions on how to achieve this?

I have attached the Waukesha projected TIFF images i used, corresponding PNG images for deeplearing and the FME-workbench for better understanding.

1 Attachments
daveatsafe
Safer
Forum|alt.badge.img+19
  • daveatsafeSafer
  • Safer
  • Best Answer
  • June 28, 2023
a_afiaax wrote:

Hello Daveatsafe,

Thank you for your helpful solution and the attached FME-workbench.

However, I encountered an issue where the offset-transformer moved all polygons to (0,0) of the projection of all images not to the (0,-0) of every non-projected PNG image, and they are overlapping each other , as shown in this screenshot:

image_00 

Each image is 2000 pixels * -2000 pixels, and each polygon corresponds to a specific house on an image, as shown in this screenshot:

image_01 

Before extracting the pixel coordinates of every polygon, I need to offset the projected polygons of every image to the right place in the their associated non-projected PNG image (2000 pix * -2000 pix and 0,-0 origin for every PNG), as shown in second screenshot.

 

Do you have any suggestions on how to achieve this?

I have attached the Waukesha projected TIFF images i used, corresponding PNG images for deeplearing and the FME-workbench for better understanding.

Having the GeoTIFFS is a great help, since they contain the georeferencing needed to align to the polygons. I am attaching a workspace that will extract the origin and pixel size from the GeoTIFFs, then use this info to move and scale the polygons to align to the pixel space coordinate.

a_afiaax
Contributor
Forum|alt.badge.img+2
  • a_afiaaxContributor
  • Author
  • Contributor
  • July 3, 2023
daveatsafe wrote:

Having the GeoTIFFS is a great help, since they contain the georeferencing needed to align to the polygons. I am attaching a workspace that will extract the origin and pixel size from the GeoTIFFs, then use this info to move and scale the polygons to align to the pixel space coordinate.

Hi @daveatsafe​, I wanted to express my gratitude for your helpful solution and the attached FME-workbench. It worked well.

Reply


Helpful Members This Week

Community Stats

31,300
Posts
118,696
Replies
38,880
Members

Latest FME

Terms & ConditionsCookie settings

Cookie policy

We use cookies to enhance and personalize your experience. If you accept you agree to our full cookie policy. Learn more about our cookies.

 
Cookie settings
A product of
Safe Software respectfully acknowledges that we live, learn and work on the traditional and unceded territories of the Kwantlen, Katzie, and Semiahmoo First Nations.