ABSTRACT
Part I Homogeneous Partitioning of the Surveillance Volume discusses the
implementation of the first of three sequentially complementary approaches for
increasing the probability of target detection within at least some of the cells of
the surveillance volume for a spatially nonGaussian or Gaussian “noise”
environment that is temporally Gaussian. This approach, identified in the Preface
as Approach A, partitions the surveillance volume into homogeneous contiguous
subdivisions. A homogeneous subdivision is one that can be subdivided into
arbitrary subgroups, each of at least 100 contiguous cells, such that all the
subgroups contain stochastic spatio-temporal “noise” sharing the same
probability density function (PDF). At least one hundred cells per subgroup are
necessary for sufficient confidence levels (see Section 4.3). The constant false-
alarm rate (CFAR) method reduces to Approach A if the CFAR “reference” cells
are within the same homogeneous subdivision as the target cell. When the noise
environment is not known a priori, then it is necessary to sample the
environment, classify and index the homogeneous subdivisions, and exclude
samples that are not homogeneous within a subdivision. If this sampling is not
done in a statistically correct manner, then Approach A can yield disappointing
results because the estimated PDF is not the actual PDF. Part I addresses this
issue.