Stereo Photogrammetry –
How We Create 3D Models
Many end users are familiar with working with a high-resolution orthoimage as a base layer for their 3D data set
so they can easily compare the model to the real world.
How aerial or satellite imagery are used to create 3D terrain and building models is generally less understood.
In this brief introduction we'll describe the key concepts and steps in this process – a set of mathematical principles and techniques
nearly as old as photography itself called stereo photogrammetry.
At its simplest, stereo photogrammetry refers to the calculation of height values (e.g., z values) by contrasting two overlapping images.
The difference in perceived location in the corresponding points in the images based on the point of view from which the image was taken is called parallax.
Using the geometry based on the intersection of two rays created from their respective image's point of view, it is possible to calculate the appropriate height values for the overlapping image area.
With frame camera aerial imagery or fixed-telescope satellite imagery, the height accuracy is determined by the base-to-height ratio (B/H ratio)
of the overlapping imagery (i.e., stereo pairs) collected along and between sensor flight paths.
The B/H ratio is simply the distance between image collection points divided by the height of the camera, or sensor.
A B/H ratio between 0.5 and 1.0 generally satisfies accurate stereo geometry.
For agile pushbroom satellite sensors such as WorldView‑1 or 2, or GeoEye‑1, we use other measurements to assess the potential accuracy of the stereo pair configuration:
the angle of convergence, the angle of asymmetry and the bisector elevation angle (BIE).
Of these, the angle of convergence is the most critical and describes the angle of separation between images forming a stereo pair.
Stereo accuracy will be poor if the convergence angle is not between 30-60 degrees.
So how is stereo photogrammetry used in the ComputaMaps 3D production process?
Once an area of interest (AOI) and product specification are established with a customer, we source the most recent and appropriate resolution data from our imagery vendors.
In fact, one technical advantage that ComputaMaps has developed is the ability to search the commercial image archives for pairs of standard collection satellite imagery
(i.e., single shot or "mono" imagery) that create satisfactory stereo geometry.
This has enabled us to greatly increase the global archive of potential high-resolution stereo satellite imagery and thus provide high accuracy 3D data to our clients for most urban AOIs worldwide.
Once the stereo models are set up we can begin the extraction of DTM and 3D building models.
For DTM generation we use a combination of automatic terrain extraction techniques using auto-correlation software as well as manual breakline editing by skilled photogrammetrists.
Our 3D building models are collected using our own highly optimized 3D feature extraction software suite.
Together with clutter, linear vectors and orthoimagery, the finished DTM and 3D building models compose the essential elements of our City Planner product.