Most of the radiation measurements in tradiional remote sensing above natural surfaces such as soils or vegetation canopies

consisted of onl nadir reflectance factors. Although such nadir measurements have the advantage of simplicity

the amount of information that can be inferred about the surface is limited mainly to that obtainable from wavelength discrimination and sun incidence angle changes. The result ofrecent studies suggests that multiple view angle may fee used to great advantage in estimating the structure and characteristics of vegetation. This advantage of off-nadir measurements is only beginning to be understood.The major achievements made in this area have been summarized in this paper

and the vegetation canopy reflectance models are reviewed according to the following four categories: geometrical optical models

turbid medium models

hybrid models and computer simulation method. The geometrical models are those that based on geometrical optical theory and the description of canopy/crown contours by some common geometrical patterns like cylinders

spheres

cones and ellipsoids etc. These models are generally applicable to inhomogeneous plant canopies

especially to the sparse canopies. The turbid medium models are developed from the solution of the radiation transfer equation in turbid medium through specifying the canopy elements by small absorbing and scattering particles （plates

discs） with finite size and given spatial orientation and optical properties

and randomly distributed in horizontal layers. This kind of models are prevalently used for the dense

horizontally uniform plant stands （crops

grasses）. The. hybrid models

which are deduced from the combination of the above two kind of models

can represent homogeneous as well in homogeneous canopies. Therefore

hybrid models are expected to have much wider ranges of application. The last kind-computer simulation method can realistically simulate the radiation interaction processes within canopies

and their dependence on various canopy structural parameters.For practical purposes

however

a certain balance between model precision and its computational efficiency is needed since precise models/methods are

as a rule

computationally tedious. It is suggested from the foregoing discussions that comprehensive analytical models may be most useful for extracting the vegetation structural parameters from the canopy reflectance spectrum. Based on the investigation of characteristics of scene directional reflectance spectrum

selection of wavebands being used and inversion of BRDF models

the main processes to quantitatively retrieve plant architectural parameters are deeply discussed. Finally

the approaches to bring the potential advantages of off-nadir observations into effect are described.