ON THE CLASSIFICATION OF TREE SYSTEMS USING SPECTRAL MULTI-ANGULAR INFORMATION

Multi-angular spectral information has been used to increase precision in classifying crops and natural vegetation. These classification systems now use only spectral information with a scheme of multivariate analysis or of decision trees, among others. In this paper, the characterization schemes of multiangular spectral information associated with vegetation and its use in classification systems are discussed. To review the feasibility of using multi-angular spectral data, a maquette-type experiment was designed for tree systems with five forest species. Measurements taken were associated with the bidirectional reflectance distribution function (BRDF). The multi-angular information was modeled in a compact form and used to define a general parameter, the slope, which comprises all the angular variation of the reflectances. The results show that, with the vegetation background remaining fixed, it is possible to discriminate tree systems, and when the vegetation background varies and there is little canopy cover, confusion arises but decreases in the measure that canopy cover increases.