The Fuel3D scanner brings handheld 3D imaging technology, originally developed for high-end medical applications, into the professional and consumer marketplace at a dramatically lower cost than comparable solutions. Whether you are a 3D printing enthusiast, game developer, designer, artist or maker, the Fuel3D scanner can help you easily capture, archive and manipulate detailed, full color 3D images. Fuel3D has three components:
- A handheld camera unit which connects to your PC or Mac via a USB connection and a separate power source
- Fuel3D Studio software, which runs which runs on your PC or Mac and works in conjunction with the scanner
- Optical targets, which are used to allow the scanner to track its motion during image capture
Working together, these components allow rapid capture of color 3D images, which can then be edited and exported into a number of different 3D formats, including .STL, .OBJ and .PLY to enable 3D printing, importing textured images into games and full color onscreen rendering.
So how did we manage to achieve this? Well, it wasn’t easy and took a lot of sweat and some long days. So read on to find out more about how the Fuel3D scanner has been developed to help you fire up your creativity!
How does the Fuel3D scanner actually work?
When taking an image, the Fuel3D scanner rapidly acquires a series of stereoscopic 2D photographs with several lighting directions. These are then processed by software to resolve a single 3D image. Under the hood, the scanner combines a number of image processing technologies to allow on-the-spot acquisition of high quality 3D images:
- Photometric imaging is used to acquire color and high-frequency 3D detail from the subject
- Optical localization is used to determine the position of the imaging device during the acquisition process
- Geometric imaging is used to acquire accurate underlying 3D shape information from the subject
- Data fusion is performed to combine the data output of the photometric and geometric processes to produce a single 3D image
- Photometric imaging for high-resolution surface detail
Photometry is the science of the measurement of light. The principle behind photometric imaging is that the image of a subject observed by a camera depends on both the shape and material properties of the subject, and the lighting conditions under which it is illuminated.
When carrying out photometric imaging, several images are taken of the subject illuminated by a single dominant light source from a number of different directions. Image processing techniques examine how the observed illumination levels across the subject vary with the change in lighting direction, calculating the direction of the “normal” to the surface of the subject or each pixel in the image, alongside maps of reflectivity, such as color. The resulting “normal map” is then integrated to provide a highly detailed 3D “range map” of the surface.
Traditional photometric systems are bulky, requiring many light sources and careful calibration. They are not portable and are generally regarded as inappropriate for capture of live subjects. A key component of the Fuel3D technology is the optical target.
Optical localization to track movement during acquisition
The Fuel3D system is handheld, and so moves during acquisition of the subject. By placing a simple optical target next to or (in the case of live subjects) onto the subject, parameters relating to this movement can be resolved. The principle behind this is that the system knows the size and layout of the optical target, and by looking for the target in the image, the scanner can accurately estimate the relative position and orientation of the scanner with respect to the target.
The target allows the Fuel3D software to calculate theposition of the camera and the light for each image. This eliminates the need for many of the functions of traditional system calibration and allows a practical hand-held system to be produced, which can compensate for small movements during photometric imaging.
The optical target allows accurate measurements (within limitations of resolution and noise) to be recovered in the X and Y dimensions of the 3D data output from photometric imaging. Z data will be subject to a degree of low-frequency distortion, and thus is not highly accurate. This is a fundamental limitation of photometric imaging techniques, which is overcome by incorporating geometric 3D imaging.
Geometric imaging for accurate underlying shape
Geometric 3D imaging resolves depth using optical triangulation, which involves resolving distance from parallax. With the Fuel3D system this is achieved by using stereoscopic imaging (two cameras and lenses) to acquire a matched pair of images of the subject, then identifying and correlating the location of features between the two images to sub-pixel accuracy. For this to be possible the subject must have a degree of random surface texture, either from variation in color, or from having a rough or wrinkled surface. The output from the geometric imaging technique is a 2D range map analogous to that provided by photometric imaging, with better underlying accuracy but lower resolution. Geometric 3D imaging gives accurate measurements of bulk shape in all three dimensions.
Data fusion to combine geometric and photometric data
The Fuel3D software incorporates proprietary algorithms to combine the data from its photometric and geometric 3D imaging systems to produce a single 3D model that is both accurate and has high resolution of surface detail. In essence, the high-accuracy, lowresolution geometric 3D data is used as a skeleton on which the higher resolution photometric 3D data is overlaid. The resulting 3D images consist of a large number (several hundred thousand) of samples, each having XYZ geometry (surface location in millimeters) and material properties (color) in 8 bit RGB.
Original of a plaster portrait reproduction Stitched 360° 3D scan of a sculpture Showing details and color Display of full 3D wireframe
What is the resolution of the Fuel3D scanner?
Output resolution from the Fuel3D scanner varies with the distance of the system to the subject. On average, the scanner has an average resolution of approximately 350 microns, with approximately 375,000 vertices and 750,000 polygons captured on a flat surface in a single scan image.
What kinds of things can be imaged with Fuel3D?
The Fuel3D scanner is a combination of hardware and software which can resolve 3D information from subjects, whether people or inanimate objects. It is important to understand that there are some scenarios in which the system will resolve data better than in others, and also some ways to maximize the effectiveness of the system.
Ideal characteristics of objects that will scan well:
Gently contoured surface with soft curves
- Varying color
- Matte, non-shiny surface
- Plants and trees
Objects with the following characteristics may not be suitable for scanning, or may require technical workarounds to capture successfully:
Cavities or protrusions
The scanner can only image what it ‘sees’ from its view point and cannot capture part of an object that is hidden behind another.
The object will absorb light from the flashes which lowers the amount of detail.
Mono-color with no texture
Objects without variation in color do not provide surface information for the scanner to work
Reflective or shiny
Objects where light reflects off the surface prevent accurate surface measurement.
Objects where light transmits through the surface prevent accurate surface measurement.
Sharp edges and corners
The scanner will trend to smooth objects with geometric features with flat surface and sharp corners.
Key points to note:
- The Fuel3D scanner has fixed focus. It resolves a maximum size of about 40cm/16inches diagonal in a single image, approximately the size of a sheet of letter sized/A4 paper.
- Textured objects are better, ideally randomly. This is because the system has to correlate pixels between stereoscopic views of the subject. Textured can mean one or both of having varying color, or surface roughness meaning illumination from an angle (as with the Fuel3D scanner) results in surface detail being visible.
- The scanner can only resolve what it can see from a single viewpoint. Acquiring, for example, a statue's head, will require multiple shots to be stitched together using the Fuel3D Studio software.
Full color 3D scan of a sunflower Hybrid 3D scan of a soft toy
(color image, relief and wireframe)
3D wireframe of a face