The objective of this work is to study the use of 3-D curves in model based object recognition. We approach the two main problems of object recognition, i.e., model formation and matching in a unified way. We propose a framework in which 3-D curves will be used both to represent objects in a database of models, and then present algorithms that use these curves to perform efficient matching between an observed object and a previously prepared database of object models. The motivation for this work comes from the fact that 3-D curves can describe in a natural way the objects from which they were extracted. Moreover, the use of these curves in the matching process has proved to be highly accurate while at the same time very efficient. In this work we present algorithms to extract 3-D curves from a pair of range and intensity images, and then algorithms that classify and separate between the different types of curves. We will also present two efficient algorithms for matching 3-D curves.