Then, to each image has been applied the supervised Maximum Likelihood Classification
(MLC) algorithm, more suitable when each class defined has a Gaussian distribution
[Bolstad and Lillesand 1991]. Finally, a 3×3 majority filter has been applied to the classified
LC data, to reduce the “salt&pepper” effect [Lillesand and Kiefer, 999].
In order to evaluate the user’s and the producer’s accuracy, a confusion matrix was applied
to the classified images [Congalton, 1991; Congalton and Green, 2009]. In particular, for
each Landsat image, the LC class assigned to 256 pixels (selected using a stratified random
sample) was visual compared with the equivalent area in the aerial frames (IGMI photos
and/or NCP orthophotos) closer to the same period. The overall accuracy values of each
classified image are reported in Table 3.
Then, to each image has been applied the supervised Maximum Likelihood Classification
(MLC) algorithm, more suitable when each class defined has a Gaussian distribution
[Bolstad and Lillesand 1991]. Finally, a 3×3 majority filter has been applied to the classified
LC data, to reduce the “salt&pepper” effect [Lillesand and Kiefer, 999].
In order to evaluate the user’s and the producer’s accuracy, a confusion matrix was applied
to the classified images [Congalton, 1991; Congalton and Green, 2009]. In particular, for
each Landsat image, the LC class assigned to 256 pixels (selected using a stratified random
sample) was visual compared with the equivalent area in the aerial frames (IGMI photos
and/or NCP orthophotos) closer to the same period. The overall accuracy values of each
classified image are reported in Table 3.
การแปล กรุณารอสักครู่..
