Lab 4
Masking clouds, shadows and unwanted
areas from a multispectral satellite image
Masking operation on raster layers in Raster Calculator
• Masking a satellite image can be performed through
combination of Set Null,~ (Boolean Not), and IsNull
• Setnull(~(ISNULL(erase layer)), orig raster)
• The Set Null tool sets identified cell locations to NoData based
on a specified criterion
• The ~ (Boolean Not) will perform a Boolean Not operation
when the input (operand) is a raster
• The IsNull returns a value of 1 if the input value is NoData and
0 for cells that are not
Batch processing for masking operation
6
1
4
3
Lab 5
Unsupervised Image
classification
Unsupervised image classification
• The user specifies the number of classes and the spectral classes are
created solely based on the numerical information in the data (i.e.
the pixel values for each of the bands or indices).
• Clustering algorithms are used to determine the natural, statistical
grouping of the data.
• The pixels are grouped together into based on their spectral
similarity: spectral clusters
• A class corresponds to a meaningful grouping of locations,
such as forests, residential areas, or water bodies, while clusters
are naturally occurring groupings in the data based on the
statistics of their attributes.
• The name of the algorithm used for creating clusters in an
unsupervised classification is Iso Cluster.
• The Iso prefix stands for Iterative Self Organizing (ISO), a
method of performing clustering.
Lab 6
Supervised Image classification
Supervised image classification
1. Data exploration and preprocessing
• Principal component analysis
• Creating a multiband image
• Creating a subset of bands for the classification
2. Collecting training samples
3. Evaluating training samples
• to explore the spectral characteristics of different areas.
• to evaluate training samples to see if there is enough
separation between the classes.
• Creating the signature file (Supervised
classification)
• A signature is a subset of cells that are
representative of a class or cluster.
• The statistics of signatures are stored in a
signature file that will be used to classify all
cells in the intersection of the input bands.
• Executing the Maximum Likelihood
Classification
or
• Executing the Interactive Supervised
Classification tool
Post-classification processing
• Post-classification processing refers to the process of removing
the noise and improving the quality of the classified output.
• some misclassified isolated pixels or small regions of pixels may
exist which gives the output a "salt and pepper" or speckled
appearance
• Majority Filter: removes the isolated pixels or noise from the classified
output.
• Boundary Clean : smooths the ragged class boundaries and clumps the
classes
Lab 7
Classification Accuracy Assessment in
ArcGIS
What is accuracy?
• Accuracy:
• Accuracy is about how close a reading is to
the actual value
• Accuracy is an expression of the lack of
error.
• Error is the difference between a
measurement and the true value of the
measurand (the quantity being measured).
• Error does not include mistakes (e.g. values
that result from reading the wrong value).
• During repeated measurements none of the
results can be preferred over the others.
• Precision – How close several readings are to
one another.
Critical Steps in Accuracy Assessment
• Determine the total number of samples to be collected for each category
A generally accepted rule of thumb: a minimum of 50 samples for each LC category.
If the area is especially large or the classification has a large number of LC categories
(i.e., more than 12 categories), the minimum number of samples should be increased to
75 to 100 samples per category.
It also depends on the relative importance of that category within the objectives of the
mapping or on the inherent variability within each of the categories.
• Design proper sampling scheme: simple random sampling, stratified, and systematic
sampling
Simple random sampling tends to undersample small but possibly very important areas
unless the sample size is significantly increased.
stratified random sampling is recommended where a minimum number of samples are
selected from each strata (i.e., category).
• Obtain ground reference information at sample locations
• Field data collection OR High Resolution Reference Imagery
• Produce error matrix
• Evaluate the error matrices
Evaluation of Error Matrices
Overall accuracy = the total number of correctly classified pixels / the total number of pixels in the error matrix (N)
Producer’s accuracy = the probability of a reference pixel being correctly classified and is a measure of omission error.
User’s accuracy = It tells use how often the class on the map will actually be present on the ground.
Comparing User's and Producer's Accuracy
• Therefore, although the producer of this
map can claim that 87% of the time an
area that was deciduous on the ground
was identified as such on the map, a user
of this map will find that only 57% of the
time that the map says an area is
deciduous will it actually be deciduous on
the ground.
• The high producer’s accuracy occurs
because too much of the map is labeled
deciduous.
• there is significant confusion in
discriminating deciduous from barren and
shrub.
Assessing the Accuracy of a Binary Map
Accuracy assessment in ArcGIS
• Step1: Create accuracy assessment
points
• Input: classification result
• Step 2: Collect reference data from High
3
resolution satellite image
• Input: Accuracy assessment points, and 1
Google Earth Image
• Step 3: Compute confusion matrix
• Input: Accuracy assessment point with its
attribute filled out
Thank you for your attention
