Science & Research

Professor, grad students make 3D scanning accessible, accurate

New algorithm synchronizes projector with detached camera to produce 3D models from ordinary devices

By
Staff Writer
Tuesday, February 2, 2016

Gabriel Taubin, a professor at the School of Engineering, and Daniel Moreno GS, a fifth-year computer engineering PhD student, recently designed an algorithm allowing ordinary smart phones and digital cameras to become 3D scanners.

The idea came to them when they started questioning if it was possible to create 3D models without synchronizing a camera and a projector — the method previously believed to be the only way to create 3D models, Moreno said.

3D scanning copies a shape from the real world onto a computer to create a 3D model, Moreno added.

“What we’ve tried to do is totally disconnect the projector from the camera,” Moreno said.

This past June, at the CVPR, an annual computer vision event in Boston, Moreno presented another method using the Projector-Camera Calibration software, which includes a camera and a projector connected to a computer.

Taubin and Moreno, along with Kilho Son GS, created the model in which a simple projector displays sequencing patterns on the object being scanned. For every pattern projected, a picture of the object is taken and then transferred onto the computer. It is necessary to synchronize the projection of each pattern to the capture of its respective image; otherwise, the image’s pattern will be blurry.

The importance of this model is its precision, Moreno said. The synchronized 3D scanner creates high-definition models with a small number of images, he added. This particular method requires a minimum of five photos to be taken — the more photos taken, the more accurate the 3D model of the object will be.

Though the items that can be scanned with this method are endless, there is a limitation to it — the camera and the projector have to be connected and synchronized.

“This is the main problem we had to solve, because (if) there is no synchronization between the projector and the camera, they are working at different speeds,” Moreno said.

If the two are not synchronized, the camera will take a photograph at a different time than when a pattern is projected, or an image will be captured while a pattern is switching to the next. The image, as a result, would display a combination of the two patterns, Moreno said.

“We had to find a way to go around that limitation, and that’s what this (new) method is about,” Taubin said.

With this newly discovered algorithm, a projector is able to run by itself and continuously project patterns, like a sequence, but the camera is completely detached, Taubin said. In this case, a smart phone or a digital camera would be able to become a part of this 3D scanning revolution, he added.

The algorithm “takes the images and calculates the new images, which are the ones that would be captured if the camera and projector were synchronized,” Moreno said. It is as if the new images are synchronized, yet the system, the camera and projector are not synchronized. With the newly synchronized images, the previous method that already existed could be used to create 3D models, either on the computer or on a smart phone, he said.

“That’s basically the novelty of this method — that we can take unsynchronized images and create synchronized ones,” Taubin said. Because of the new algorithm, cell phones and digital cameras are able to take photographs that will become 3D models.

Moreno and Taubin plan on building a mini projector about the size of a flash drive that will project patterns in sequence, Moreno said. They also plan on making a smart phone app that will run the software for the 3D scanning. “You could potentially build a model, send it to the printer and make a copy,” Taubin said.

This 3D scanning method is the first of its kind. Engineers and industrial designers use expensive 3D scanners that cost well over $50,000, to which small industries do not have access, Taubin said. On the other hand, devices such as Microsoft Kinect, which was designed to track the motion of people for games, are used as 3D scanners but are not of the same quality, he added.

Taubin and Moreno are also trying to do “something in the middle — good quality and low cost,” Taubin said. The University’s Technology Ventures Office has been helping them file patents for the device that projects the patterns as well as the software that will become a smart phone application. The two are currently in the early stages of exploring the possibility of commercializing their method, he added.