Science

Pix4D creates 3D images from 2D aerial photos

Pix4D creates 3D images from 2D aerial photos
Pix4D is a program that creates 3D aerial images by combining hundreds of 2D photographs, shot by an unmanned aerial vehicle (Image: Pix4D)
Pix4D is a program that creates 3D aerial images by combining hundreds of 2D photographs, shot by an unmanned aerial vehicle (Image: Pix4D)
View 1 Image
Pix4D is a program that creates 3D aerial images by combining hundreds of 2D photographs, shot by an unmanned aerial vehicle (Image: Pix4D)
1/1
Pix4D is a program that creates 3D aerial images by combining hundreds of 2D photographs, shot by an unmanned aerial vehicle (Image: Pix4D)

While Google Earth can be extremely useful - not to mention a lot of fun - it now has some competition in the form of Pix4D. Instead of satellites, the imaging system uses a small, relatively inexpensive unmanned aerial vehicle (UAV) to acquire several hundred 2D photographs of a given geographical area. Those photos are then merged into one image, which users can explore in three dimensions on a computer screen.

The UAV (which the user has to provide) flies back and forth across the town, field, or other area, taking pictures as it goes - essentially, it's acting as an airborne scanner. The still images it collects are then loaded onto the Pix4D software. The system automatically detects "interest points," which are items with high visual contrast, or that are otherwise visually distinct. By comparing the orientation of these same interest points in photos taken from a variety of angles, it is able to build a three-dimensional photographic representation of that item, along with everything around it.

It's the same principle that allows our brains to take slightly differently-angled images of the same scene from each of our eyes, and combine them into one 3D image.

Pix4D then proceeds to build a sparse 3D model of the entire area by combining all the images, using GPS tags on each photo as a guide. That's followed by a dense 3D model, to which texturing is finally added. The entire process (not counting the initial fly-over) varies with the size of the area, although in one example, 412 photos were combined into one 3D aerial image in less than one hour.

Users can choose between a high-speed cloud-based service, a private server based within a network, or a light version of the program that can run on a laptop.

The technology was developed at Switzerland's Ecole Polytechnique Fédérale de Lausanne, and the Pix4D company is a spin-off of that project. Below is a video of a 3D image of the town of Lausanne, created using the technology.

Source: New Scientist

Lausanne dense 3D model

2 comments
2 comments
Brillig
This reminds me of a demonstration I saw in 1984 (yeah, 27 years ago). Pixar had a display at Siggraph in Dallas. They took photos of Mars taken by Viking, and their computer analyzed the various pictures to create a model of the Martian surface. Then, they animated a simulated helicopter ride over the surface of Mars. This was all done in just a few minutes.
Lamar Havard
I tried a technique where you make two images and place them side-by-side and cross your eyes. You\'ll see three pics, but the middle one, because your eyes are combining two pics in front of each eye, will be true 3D. It WORKED!