The land cover information of Berkeley was derived from natural
color aerial photographs provided by the California Spatial
Information Library. The one meter ground sample distance (GSD)
orthophotos were acquired and rectified by the National Agriculture
Imagery Program (NAIP) in 2005. The entire area of Berkeley
was covered in four scenes. Those four scenes were mosaiced into
one image and the portion of Berkeley City was cut out for classification.
The water surface was manually delineated from the
image and masked out. After that, an object-based image analysis
(OBIA) approach was used to classify the image into different
land cover classes. The OBIA is more suitable for classifying high
spatial resolution imagery for urban environments than other classification
methods (Cleve et al., 2008; Jacquin et al., 2008). In this
study the image was first segmented into image objects using a
multi-resolution segmentation algorithm. After segmentation, an
iterative classification process was run to classify image objects
into land cover classes until the classification results could not be
improved by more iteration.