Recognizing Three-Dimensional Objects By Their Silhouettes

Abstract

We describe a method of classifying or identifying a three-dimensional object from one or more of its silhouettes. The method is based on a low-cost parallel mechanism for computing the slope density, i.e., the density of slopes of the edge of the silhouette. Examples of applications of the method are microscopy and aerial navigation. In both of these applications the object's outline is often the principal clue to its identity, even though other clues, such as texture, color and interior detail may be available. The present study is based on an application of our technique to recognizing aircraft.