Berkeley was initially classified i

Berkeley was initially classified into six land cover classes. Dark
impervious surface included asphalt surfaces, colored roofs, and old
concrete surfaces. Light impervious surface included light colored
roofs and newconcrete surfaces. Tree and shrub coverwas the area
covered by the perpendicular projection of the canopies of trees
and shrubs. Grass included both managed lawn and unmanaged
grasses. The other two land cover classes were water and bare soil.
In post-processing, the two impervious surfaces were merged into
one class: impervious surfaces. Spectral and texture features of the
image selected through trial tests were used in classification. Spectral
features including the mean value of blue channel, mean value
of red channel, and a customized feature, the mean value of green
channel divided by the sum of mean values of green channel and
red channel.We found that this customized feature separated vegetation
from other land cover types very well. Two texture features
were also used in classification: the gray level co-occurrence matrix
(GLCM) contrast and GLCM entropy. GLCM contrast measures the
amount of local variation of gray levels of pixels in the image; a
high value means less homogeneity. GLCM entropy measures the