An Introduction to Remotely Sensed Data Analysis - Sampling Spatial Units for Agricultural Surveys - page 88

covariance matrices, generated by i.gensig or i.class based on regions of
pixels, to determine the category to which each cell most likely belongs.
GRASS also provides another two-step method for supervised classification.
The first step of this method is executed through the command i.gensigset .It
is a non-interactive method that reads the training map. Then, i.gensigset
extracts spectral statistics from an image using the classification of the pixels in the
training map, and makes these statistics available to i.smap .The i.smap
program segments multispectral images using a spectral class model known as a
Gaussian mixture distribution. i.smap has two modes of operation. The first mode
is the sequential maximum a posteriori (SMAP) model, which takes into account
the similarities of pixels in the neighborhood. The SMAP segmentation algorithm
attempts to improve the segmentation accuracy by splitting the image into regions
rather than separately segmenting each pixel. The second mode is the more
conventional maximum likelihood classification, which classifies each pixel sepa-
rately but requires less computation.
GRASS provides analysis for time series processing ( r.series ). Statistics can
be derived from multi-temporal satellite data. The common univariate statistics and
linear regressions can also be calculated.
Conclusion
The purpose of this chapter was to present some basic concepts for the
analysis of RS images. We have examined some procedures for restoring
images and for extracting thematic information.
Recently, satellite and/or aerial RS technology combined with in-situ
observations has become an important technique for improving the present
systems of acquiring and generating agricultural and resource data. To benefit
from remotely sensed data, managers, consultants, and technicians must
understand and to be able to interpret the images.
RS has been increasingly considered for developing standardized, faster,
and possibly cheaper methods for agricultural statistics. Many countries have
RS programs providing data to official agricultural statistics programs.
Carfagna and Gallego ( 2005 ) provided an exhaustive description of the
different uses of RS for agricultural statistics.
RS techniques can represent an appropriate support for particular problems
in agricultural surveys such as data reliability, incomplete sample frame and
sample size, unit selection, area measurement, non-sampling errors, gaps in
geographical coverage, and non-availability of statistics at a disaggregated level.
RS can be appropriately used at the design level (see Chap. 6 ) . Remotely
sensed images provide an overview of the area under investigation, and are
useful when constructing the spatial reference frame (see Chap. 5 ). For an
empirical example, see also Sect. 2.3.3 , where we describe the Italian AGRIT
program and the use of RS for defining the survey frame.
(continued)