We present a new measurement technique based on image correlation to capture the wind-induced motions of crop canopies. Tests were carried out on an alfalfa field submitted to natural wind. The motion is video-recorded from the side of the field and we take advantage of the natural periodicity of the crop to correct the distortion due to perspective. The computation of the canopy motions between successive images is based on Particle Image Velocimetry (PIV), and takes advantage of the natural small-scale heterogeneities of the canopy. It provides the two-dimensionnal spatio-temporal velocity field of the crop top surface with high spatial and temporal resolution. The Bi-Orthogonal Decomposition (BOD) of the velocity field then reveals large coherent propagating structures that scale with typical wavelength of wind fluctuation over canopies.