We are now witnessing the emergence of 3D facial recognition. The results published at the end of evaluation campaigns as FRVT (Facial Recognition Vendor Test 2000/2002) showed that facial recognition approaches suffer from several problems related to changes in pose, lighting and occlusion,. 3D method was then proposed to overcome problems related to the pose and lighting. Nevertheless, taking into account this new method requires adapted processing algorithms and eventually to fuse 2D and 3D method in recognition stage. Different fusion strategies used in literature can be divided into three categories: complementary fusion, competitive fusion and cascade fusion. The complementary fusion consists in a combination of facial features extracted from both 2D and 3D to construct a single feature vector. The competitive fusion consists in combining the similarity scores generated by different classifiers for a single score. Cascade fusion architecture admit a series of different classifiers. All these methods are independent. Therefore, we plan to develop an innovative method that brings together all these algorithms in one system. Subsequently, we try to calculate the optimal fusion strategy. This strategy ensures: the selection of the best features, selecting the best classifiers and selecting the best fusion method. We intend to build on the results of our work to study the effectiveness of the models obtained in a real identification scheme. Finally, experiments performed on a dedicated 3D face database like FRGC 2.0 allow us to evaluate our methods.