Photogrammetry, point clouds and stained glass

Last weekend I met with Jules Mominee of Mominee Studios [nationally renowned designers of fine stained glass and restorers of historic art glass] to conduct a work shop demonstrating how photogrammetry can add value to his work. We visited Trinity Episcopal Church in Staunton, VA to choose a test subject from their rich collection of stained glass windows – and selected the triptych behind the allar which was designed by the celebrated Tiffany Glass and Decorating Co. of New York in 1897.

My goal was to demonstrate how photographing the windows with a calibrated camera+lens combination could produce a valuable documentary record of these important heirlooms above and beyond standard photography. I would show how we could use the photographs as the basis for rectified scale-able photographs (with all lens and parallax distortion removed). I also wanted to show how we could “go into” the photographs and extract precise 3D point measurements as needed to create measured drawings and such.

This blog post will try to cover what we did.

Photographing the triptych with telescoping tripod. Note surveyor’s rod (to establish real world distance in the photos) and a white balance target.

The first step involved shooting overlapping photographs of the subject with a different lenses. Some of the shots captured the scene in its entirety while others captured  smaller regions in greater detail (for use later as pieces of a mosaic).

The variety of images shot loaded into photogrammetric software

Next up, we processed the photographs using software that automatically calculates the relationship of the camera stations to one another and creates a point cloud describing features in common captured by the photographs.

Point Cloud representing the stained glass (in true color) and the relative 3D locations for each photograph used.

The point cloud is essentially flat (due to what it is depicting) but nonetheless consists of an agglomeration of precise 3D measurements. Here is an animation showing its three dimensional nature:

An animation showing the point cloud depicting the stained glass triptych and the camera stations (in red)

Then we chose a handful of “smart points” relating to specific locations on the glass in order to establish a meaningful coordinate system. These points are shown on the images below.

Location of “smart points” on center window (lower portion)

Location of “smart points” on center window (upper portion)

Location of “smart points” on right window

Once these ponts were chosen and used to define our principal plane, we recalculated the model (with “smart points” on our surveying rod to establish real world dimensions).  Here are the x, y, and z values for our smart points:

Object Point Calculation Table

If you look closely at these values you’ll find that the average error value for this small batch of points is calculated to be about one one hundredth of an inch. Photogrammetric analysis (esp. when using controls and targets) can greatly exceed this level of accuracy – but this is already well beyond what would be required to replicate this design.

On to image rectification… The next step is to use these same 3D coordinates to define theoretical planes onto which the individual photographs will be projected so that the resulting images match precisely the real world conditions of the glass surface.

Defining a rectification plane with four or more points

The window above shows a plane formed by points 5, 6, 7 and 8 that has a maximum error value of about a sixteenth of an inch (which means that this portion of the window is pretty flat – if there were buckling and such, as will happen with windows over a hundred years old, this value would be greater…). So this will be the spatial plane onto which we will rectify the photo of the center window, lower portion.

Next up, we made a lasso of the area of the photo that we want to rectify since not all of the image corresponds to our rectification plane.

Lasso indicating extent of image that is coplanar to the rectification plane.

Then we were ready to create our rectified image of the triptych in its entirety by creating a mosaic of four smaller rectified images. In the way that we shot this example, we were able to create a rectified image that would respect the graphic scale of 3″=1′-0″ (1:4) when printed at 150 dots per inch. This ‘resolution’ can be increased by shooting more images that are in closer range to the surface being documented.

Creating a mosaic of individually rectified regions

And here is an overview of our finished result:

with some additional images “zoomed in”:

100% crop

400% crop

So this is the level of detail available across the entire surface of the three windows. If the image were printed on several sheets full size, we would produce, effectively, the same type of document as if we were able to make a high quality “rubbing” of the window – with out having to remove it and in a fraction of the time (and in color!!!)

Continuing on, I showed how the image could also be brought into a CAD program  (such as AutoCAD) in order to create a highly detailed measured drawing in vector format. In this scenario one can directly query the image to get real world dimensions.

Overview of Triptych in CAD

Closer up view in CAD

Detail view in CAD

So, in the end we showed the value of photogrammetry as a high quality AND cost effective tool for documenting heritage artifacts such as stained glass both for restoration purposes as well as for insurance purpose to provide a reliable document in the event of catastrophic loss. It also can provide a way to share the unmatched artistry of these windows to any who would like to have a closer look.

.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s