Tag Archives: scan

Overlapping Photogrammetric systems

View of Point Cloud from side/rear

Spending a lot of time seeing how different photogrammetric systems and point cloud software packages speak to one another. sometimes the interface is elegant, sometime not so much…

Aerial Photography > Dense Point Clouds

Since 2009 I’ve done a fair amount of work with aerial photography, all the while continuing to explore architectural photography and photogrammetry while on the ground. I’ve tried in the past with some success to merge the two things but have found the results to be lacking something. That something, I think, is the hyper dense point cloud – and now I have it pretty much figured out.

Below are tow animated GIF images showing first the point cloud describing our local hospital (which I was contracted to cover with aerial photographs a few years back) and the second is a textured 3-D model of the scan.

Its so amazing to think that the flight over this structure can be reconstructed virtually a few years later with even more options to “fly around”…

Dense Point Cloud

Dense Point Cloud

Textured 3-D model

Textured 3-D model

As found drawings

"As Found"

There is a distinction between “as found” drawings (drawings that depict the reality of an existing building as it is “found” today) and drawings that were “found” in a search to locate historical documentation about a building. This is not to knock the value of original drawings that are sometimes available when working on an historic structure. Its just a reminder that original drawings do not always provide accurate descriptions of what was built. Depending on when and how the building was constructed, design drawings sometimes were more like polite suggestions.

The other issue with “found” drawings is how they are converted into CAD drawings or BIM models. Care must be taken to avoid a messy foundation for preservation work.

Test Project: Marquis Building

Angle View of Building

In between wrapping up summer projects and starting some new ones,  I wanted to work on my technique with Elcovision’s powerful new module that allows dense point clouds to be extracted from photographs.  I chose the “Marquis Building” on Beverley Street in Staunton for fairly obvious reasons. (In fact there it is at the top of this page on my mast head – from was a two dimensional rectification of the entire block)

Left Photo

(Note: I used considerably more photographs than just these three shown here)

Right Photo

Over the summer I did some training in Switzerland on how to use the new software module – but I did not have access to all of my regular equipment.  So this test project was the first time I had a chance to enhance my ‘overlap coverage’ by the use of a telescoping tripod.

The point cloud created from the analysis phase is dense enough to provide a quick sort of three dimensional model of the building even without any additional drawing.

In Elcovision – or AutoCAD for that matter – one can “hover” around the point cloud and take snapshots in either perspective or isometric views.

Plan View of Point Cloud

True Color Point Cloud

Point Cloud of Staunton Creamery Building

Here I am posting a point cloud scan of a structure much more mundane than the Arc De Triomphe, a square-ish two storey building located not far from my office.  This was produced from about 40 photographs and consists of tens of thousands of point measurements that are represented in their true color.

Staunton Creamery

Point Cloud Scan of Column Base using Photographs / Photogrammetry

Point Cloud Scan of Column Base

My second sample project using photographs to create dense point clouds to describe a non orthogonal object used a column base from the cathedral de Saint Pierre de Geneve.  These stone surfaces proved to be easier to capture than the sleek, reflective brass surfaces o the sculpture in my previous post.

This point cloud was created with a series of a half dozen photographs shot in an arc from left to right around the subject matter.  We can see here that in oreder to get all of the surfaces adequately rendered, one needs to also change the height of the camera’s point of view.

As it is, the point cloud consists of over 15,000 points. The points can be imported directly into AutoCAD with their true colors and manipulated there to be transformed into a surface model – or simply as the basis for a measured line drawing.  As for accuracy, 75% of the point measurements (which would include virtually all of the measurements describing the subject close at hand) are accurate to less than 1/8″.  Half of them are more accurate than 1/16″.

Here is a photograph of the column base, followed by two more views of the point cloud:

Column base at the Cathedral de Saint Pierre de Geneve

Plan View of the Point Cloud

Side View of the Point Cloud