A comparison is made between the knowledge base and the features extracted from the observed objects. We plan to merge existing approaches to achieve the 3D comparison. One of the most interesting techniques is the one proposed by S. Biasotti in this article. A skeletal signature of the 3D object is computed. That signature is then used as a size graph to compute discrete size functions, giving a similarity measure between shapes. A template matching algorithm will also be used. Here the idea is to generate 2D projections of 3D models in key views to be matched to observed 2D silhouettes. Template matching can be difficult because animals are deformable objects, but the system is to try both approaches and keep the stronger result. Every identification criterion we retrieve from these techniques and from our earlier observations will be taken into account in a probabilistic equation to achieve identification of the objects. The system also needs to score its confidence in the identification, so the user can verify the most uncertain identifications. At the end of this step of the algorithm, it will be determined whether the observed object is an animal and wich one in its knowledge base is the most accurate match.