The pattern recognition is always one of the most important aspectsin the remote sensing application field.As we know

the visual interpretation ofrocks is based on not only the spectral characteristics of rocks

but espiclally theirtexture structures.Most of the traditional texture analysls methods are developed byusing the statistical characteristics of the textures in the spacial and frequentialfields. In 1968

Mandelbrot first brought forth the concept of fractional Brownian mo-tion field

and later pronounced that many of the natural object:and procedurescould be well modelled bv fractional Brownian function. Based on the above theory about fBm model

we developed a method for textureanalysis

which was applied to classification of rocks in the test area selected fromMSS image of Guiyang scene

and got a satisfactory result.But

the fBm model isa long-correlative model

and is more sensitive to the low frequential informationsthan to the high ones

therefore

it is more suitdble for the rock recognition in tho- se areas where the terrain doesn’t undulate very forcely and the image textures areisotropic and well corresponding to various rocks.