3D Scanning

Custom 3D Scanning Software for the Medical Imaging Sphere

Our Partner:

Our partner, a small Venture-Capital backed company, designed and produced powerful 3-D scanning hardware, engineered a vision of their technology's role in the medical imaging sphere, but needed imaging software to complete their vision.

Although they are specialists in the medical field and were highly experienced with optical hardware, they had done few software projects, making Digital Element a perfect partner.

The Challenge:

Digital Element needed to work directly with our partner's 3D scanners- robust tools for creating 3D mesh and texture models from optical data.

One major challenge concerned the cost of the hardware system. Since the system's cost needed to be within a particular range, there were limits to the speed of the cameras and projectors. Highly efficient imaging algorithms were a must.

The second challenge was not technical, but just as important. Digital Element had to help our partner show demonstrable progress at very regular intervals, even within the first month.

The Solution:

Digital Element created a system that projected a series of images, from three sources, onto the scanned subject and, over the course of 0.3 seconds, captured  images on three separate cameras. With a low-level hardware SDK, we built the synchronization routines as well as the calibration tools.
Leveraging our proprietary depth-based renderer, Digital Element provided basic model viewing extremely early in development. Additionally, with our, 2-week sprint, Agile development approach, we efficiently communicated our constant progress throughout the whole project.

We  first employed publicly-available white papers on 3D extrusion and set the input to the software at 15 frames for each of three cameras, where each set of 15 mapped into its own voxel cloud as well as a neutral-light texture shot.

Since noise was created by refractive and reflective material in the light's path, Digital Element then developed algorithms for enhancing the voxel-cloud data. Additionally, slight movements by the scanned subject during the 0.3 second capture also created noise, requiring the construction and implementation of advanced algorithms.

The resultant thee point voxel clouds from each camera then needed to be merged to produce a 3D image. The merge was not trivial and the pressure for success was high, so Digital Element simultaneously developed three competing solutions, eventually deciding on a hybrid of two for the installed solution.

In addition, the three textures taken from the three cameras needed to be unified, with the added issue that the room's lighting, shape of the subject, and shadows could affect some camera perspectives more than others. Digital Element developed a series of algorithms to normalize the textures as well as dynamically combine best textures together.

The Results: