Gravitational lensing permits to probe the total mass distribution regardless the physical state and the nature of the (dark) matter. Weak lensing studies have more particularly focussed on the mass distribution of a dozen of high-redshift clusters (see Fort & Mellier 1994, Schneider 1996, Mellier 1999). By measuring the tidal field experienced by ray bundles of the distant background galaxies, the projected mass distribution can be recovered. The statistical properties of the shear fiels, like the rms shear or the 2-point correlation provide a direct determination of the projected power spectrum of the Large-Scale structures, P(k) (see Blandford et al 1991, Miralda-Escude 1991). The non-linear evolution of the matter density fields induces strong non-Gaussian features on scales below 10 Mpc that can be analysed from the shear statistics on scales below 1 degre (Bernardeau et al 1996, Jain & Seljak 1997, Van Waerbeke et al 1999). The wide field survey described in this proposal can thus investigate the behavior of the dark matter in these two regimes. It can be shown that the variance and the skewness of the projected mass density (inferred directly form the convergence ) provide the slope of the projected power spectrum and independently (Bernardeau et al 1996, Jain & Seljak 1997). The wide field data will also permit to detect rare events produced by massive dark halos, thus providing a unique set of shear-selected sample of clusters of galaxies and groups. Kruse & Schneider (1999) have estimated that 10 such halos per sq. degres will be detected, depending on the cosmological model. The statistical results can be directly compared to the predictions of numerical simulations. With a total area of 8° x 8° Van Waerbeke et al (1999) have show that the shape of p(k) can be recovered with about 1% accuracy on scales rangin from 2' to few degres. The cosmological parameters can also be very well recovered. For instance, a = 0.3 and a = 1 universes can be disentangle with more than 6 confidence level.