Ground-based cloud recognition plays an essential
role for automatic cloud observation. In particular, the
recognition of clouds is remarkably challenging because that
the shape, size, and composition of cloud is extremely variable
under different atmospheric conditions. A new method is
proposed to extract texturral feature using Bidimensional
Empirical Mode Decomposition(BEMD) and Tamura textural
analysis.. Cloud was decomposed into several IMFs by BEMD.
Radial basis function polynomial interpolation was applied to
construct the envelope. Then the number of zero-crossing,
means and standard deviation of the amplitude in each IMFs
were selected as the eigenvector for training processing. And
Tamura textural feature analysis was used to extract the
feature of directionality. Characteristics of the sample
database cloud was established by synthesizing the two
normalized eigenvector. The same method was applied to the
images to be identified, then the images were categorized
compared with the eigenvector of sample database by the
average sample method. The simulated experiments show that
the ground-based cloud can be recognized effectively by new
method.