Unsupervised KMeans image classification Application

Brief Description

Unsupervised KMeans image classification

Tags

Segmentation, Learning

Long Description

Performs Unsupervised KMeans image classification.

Parameters

• [param] Input Image (-in): Input image filename.

• [param] Output Image (-out): Output image filename.

• [param] Available RAM (-ram): Available RAM

• [param] Validity Mask (-vm): Validity mask. Only non-zero pixels will be used to estimate KMeans modes.

• [param] Training set size (-ts): Size of the training set.

• [param] Training set sample selection probability (-tp): Probability for a sample to be selected in the training set.

• [param] Number of classes (-nc): number of modes, which will be used to generate class membership.

• [param] Initial class centroid probability (-cp): Probability for a pixel to be selected as an initial class centroid

• [param] Number of lines for each streaming block (-sl): input image will be divided into sl lines.

Limitations

None

Authors

OTB-Team

See also

Example of use

• Parameters to set value:

• Input Image: poupees_sub.png

• Output Image: ClassificationFilterOuptut.tif

• Validity Mask: mask_KMeans.png

• Training set size: 100

• Training set sample selection probability: 0.5

• Number of classes: 5

• Initial class centroid probability: 0.9

• Number of lines for each streaming block: 100

• Command line to execute:

otbcli_KMeansClassification -in poupees_sub.png -out ClassificationFilterOuptut.tif -vm mask_KMeans.png -ts 100 -tp 0.5 -nc 5 -cp 0.9 -sl 100