The identification of mechanical parameters for real structures is still a challenge. With the improvement of optical full-field measurement techniques, it has become easier, but in spite of many publications showing the feasibility of such methods, experimental results are still scarce. In this paper we present a first step towards a global approach of mechanical identification for composite materials. The chosen mechanical test is an open-hole tensile test according to standard recommendations. For the moment, experimental data are provided by a moire interferometry setup. The global principle of the identification developed in this paper is to minimize a discrepancy between experimental and theoretical results, expressed as a cost function, using a Levenberg-Marquardt algorithm. This approach has the advantage of having high adaptability, largely because the optical system the signal processing as well as the mechanical aspects, can be taken into consideration by the model. In this paper we consider different types of cost functions, which are tested using an identifiability criterion. Although mechanically based cost functions have been studied, a simpler mathematical form is finally more efficient. Two different models were tested. The first is an analytical model based on the Lekhnitskii approach. This approach has the advantage of being a meshless solution; however, the results appeared to be partially false due to boundary effects, leading to a second approach, a classical finite element analysis. The resulting identified values are similar to values from classical mechanical tests (within 6%), which, in practice, validates our approach.