Aug 24, 2015 | By Simon

Within the past few years, low-cost object scanning has become increasingly popular thanks in no small part to the increased overall interest in 3D printing from the general public.  While creating 3D models from scratch is more often than not the desired way to create 3D content - particularly because it allows a creator to fine-tune details - 3D scanning is nonetheless a valuable way of capturing content to reproduce using a 3D printer.  

While there have been a range of new hardware devices that aim to help drop the cost of entry for those getting started in 3D scanning - such as Google’s Project Tango - there is still a lot to be desired in terms of something that’s even cheaper and perhaps already exists in our pockets - like an app for our smartphones, without the need for any additional hardware.

More recently, Peter Ondruska - a student in the Mobile Robotics Group at the University of Oxford - along with Pushmeet Kohli and Shahram Izadi of Microsoft Research have developed a new system for making real-time scans of 3D surfaces using the existing tech within our smartphones - without any hardware modifications.  

“Existing state of the art mobile methods approximate surfaces using simple visual hull constraints, Delaunay triangulation,  purely point-based representations; or use cloud-based processing to avoid on-device computation,” explains the team in their report.  

“Instead, our system reconstructs real-time 3D surface models directly on the device, allowing lightweight capture of detailed 3D objects, at speeds that have yet to be demonstrated by other systems.” 

While they spent a lot of time researching and understanding early 3D scanning systems - such as the Microsoft Kinect - the team wanted to focus on the potential for mobile phones specifically - an area that is already seeing considerable growth and is starting to include more technologies that previously weren’t possible to include on a pocket-sized device.  

The result of the team’s research and development is a new real-time volumetric surface reconstruction and dense tracking system called MobileFusion.  

In order to scan an object, the system relies on a proven five-step process: dense image alignment for pose estimation, key frame selection, dense stereo matching, volumetric update, and ray-tracing for surface extraction.  

When given a stream of image frames, the tracker estimates the camera pose by aligning the entire RGB frame with a projection of the current volumetric model.  Based on this live input, keyframes are selected using a metric that evaluates the overlap between frames.  After the overlaps have been evaluated, a camera pose is used to select a keyframe which can be used with the live frame to perform dense stereo matching.  From this stage, the output depth map is then fused into a single global volumetric model, which is displayed to the user for live feedback. 

In order to ensure that the results for each model would remain consistent during the scanning process, the system takes a hold of the devices camera settings and ensures that ISO sensitivity, exposure mode and camera focus remain constant in an effort to prevent changes in appearance and camera geometry caused by lighting changes and refocusing.  

While the developments have been promising, the team is quick to admit that their new system is not without its own unique challenges, too.  

“Obviously our method also has some limitations,” they explain.

“As with many passive stereo methods, objects with no texture cause problems. This could be fixed by using depth estimation algorithms that employ regularization to extrapolate depth values for pixels where left- right check fails. In the presence of specular or glossy surfaces, the model can contain missing geometry or noise. Color bleeding across voxels is also an issue.”

Regardless, the research is a critical step towards enabling a new generation of smartphone users to be able to take 3D scans for 3D printing just as easily as they take photographs to create their Instagrams.

“We believe these types of 3D scanning technologies will bring 3D acquisition to a new audience, where lightweight capture of the real-world becomes truly ubiquitous,” the team added.

Currently, the app is not available for download - however the team is planning on launching it soon.      

 

Posted in 3D Scanning

 

 

Maybe you also like:


   


Sam wrote at 7/27/2017 7:52:25 PM:

Why has this project not gone anywhere? It seems like the answer that the industry is missing right now.

Phil wrote at 3/17/2017 11:54:16 PM:

I'm interested in this product. When will it be available?



Leave a comment:

Your Name:

 


Subscribe us to

3ders.org Feeds 3ders.org twitter 3ders.org facebook   

About 3Ders.org

3Ders.org provides the latest news about 3D printing technology and 3D printers. We are now seven years old and have around 1.5 million unique visitors per month.

News Archive