Scans pick up object orientation

It only seems easy to recognize a familiar face -- or any other object, for that matter -- from any angle. Pattern recognition is difficult for computers, but has many potential uses in areas like manufacturing and security.

Researchers from the University of Valencia in Spain have improved the process with a method of mapping three-dimensional objects in a way that records every possible point of view of the object. Because the method produces a map that contains the information of the object from every point of view it allows objects to be recognized from a wider range of views.

The method could be used to inspect the orientation and shapes of manufactured goods, and also for face recognition, according to the researchers.

The researchers' method calls for scanning objects from every angle using a range finder that maps the peaks and valleys of the object's surfaces. A later scan from one angle can be compared to this mapped to determine the orientation of the object regardless of the object's orientation or scale, according to the researchers.

Similar methods have been able to detect object orientation from a limited number of angles, according to the researchers. The key to detecting object orientation from every angle was mathematically converting the three-dimensional scan map to a series of points on a sphere.

The method could be used practically in two to five years, according to the researchers. The work appeared in the December 15, 2003 issue of Optics Express.