Three-dimensional object recognition is a crucial problem in computer vision. It has attracted attention of many researchers in the last two decades. There are many related applications in real life, such as capsule endoscopy, cartography, hostile areas exploration, or even micro-UAVs. These applications could derive great benefit from having the opportunity to rebuild and recognize their environment in 3D, but existing systems are too limited regarding size and power consumption. Despite many methods have been proposed, effective and complete solution to this problem is still being sought. To meet certain needs of these applications, we present in this thesis the Cyclope project, an embedded active stereo-vision system that is able to give in real time both 3D data and texture, and uses Support Vector Machines SVM for objects classification. The proposed architecture meets the constraints defined in such a project: high integration level, high accuracy and real-time processing.