The recent advances in the development of UAV platforms along with the increasing performances of sensors resulted in a boom in the exploitation of this new type of observation. UAV observations in the visible to near infrared domain are characterized by extremely high resolution images (down to a fraction of mm) in relation with the low flying altitude (few tenths of m), the possibility to easily acquire directional observations and to build digital surface model using large overlaps between adjacent images. This presentation illustrates the use of RGB and multispectral cameras aboard UAV for vegetation characterization. It includes estimates of plant, flower or reproductive organs population density, the description of the 3D envelope of plants including the derivation of plant height, the retrieval of green area index, the fraction of intercepted radiation, the leaf chlorophyll content, the scoring of disease symptoms on crops and the identification of specific stages. The exploitation of the image in satellite remote sensing is mostly restricted to the use of pixel information without taking into account its neighbors. Since most objects in the landscape have typical length scales smaller than the pixel size, satellite observations provides a statistical description of the pixel content. Conversely, the extremely high resolution available from UAV observations allows to clearly identify organs or plants, and in any case describe the image texture that potentially vehicles specific information on the target. The exploitation of such imagery forces to use new methods that are generally object centered. Some examples are given to illustrate these new algorithms. Machine learning approaches will take a dominant part of the techniques to be used. It includes as well techniques that provide descriptors of the image that will be related to some variables of interest for the user. The use of physical models that incorporates a description of the biological processes involved in the development of the plant structure will also play an important role in data interpretation. Finally, the complementarity between satellite and UAV observations is addressed. Satellite such as Sentinel 2 provides a very large and systematic coverage with a revisit period of few days at a very low cost per km², while UAVs will provide very limited coverage but with much better performances to characterize the surface at the expense of a much higher cost per km² for its implementation and data processing.