How to Execute a Classification
This page describes the procedures of the "Classification"
window, used to classify an image either by pixel or by regions
( make a segmentation before).
See about image classification in the SPRING
Sequence to execute a classification
STEP 1
Creating a context file:
- activate the database and the project that has the image(s) to be classified;
- activate an image IL and draw in the active window. The image to be classified
does not need to be the image used to get samples, for this task it is even possible to use a
synthetic image;
- click on Image and select Classification...; The
"Classification" window will be presented;
- click on Directory... to inform where the context
file will be created. The "File Selection"
or "Directory Selection" is presented (see How to proceed - details). By
default SPRING suggests the directory corresponding to the active project, and this file will have
the *.ctx extension. It is recommended to keep the suggested directory, because when performing a
database backup these files will be included;
- click on Create...; The "Context Creation"
window will be presented;
- after creating the context file see proceeding for training execution.
NOTE: In case the user wants to remove a context file, click on a name in the list and next
click on Remove.
STEP 2
Training Execution:
- in the "Classification" window, after
selecting a context file from the list, click on Training...
to define the samples in the "Training"
window. The system requires any image in the active window, thus show one, in case this was not done before.
Any image can be used, even one that is not included among the ones selected from the context file;
- Once the samples are created see procedure for sample analysis and image classification.
STEP 3
Sample Analysis and Image Classification:
- in the "Classification" window, once the
training execution is done and the samples acquired, click on Classification... to analyze the window "Image Classification"
and decide to classify the image;
- once the image is classified, the user may or may not decide on executing a
post-classification.
STEP 4
Normally a pixel classification outcome is very noisy, mainly in image regions where there is an intense
variation among the gray levels. Thus, a filtering process can be useful to eliminate some isolated pixels,
but this is not mandatory.
Executing a Post-Classification:
- in the "Classification" window, once the
image classification was performed, click on Post-Classification... to execute this task. The
"Post
classification" window is presented;
- once the image post-classification was executed, the user can perform a Mapping to create a theatrical map.
STEP 5
Executing a Mapping for Thematic Classes:
See also:
Creating a context file
To Execute the Training
How to Execute a Classification
Sample Analysis
Post-classification
Thematic Classes Mapping
|