GRASS Logo

GRASS Help System


IMAGE CLASSIFICATION

One of the methods of deriving information from digital imagery is multispectral classification. Multispectral classification analyzes the spectral reflectance values of pixels in an image. Pixels, or picture elements, are the cells that make up an image. The word "pixel" is usually used in reference to a cell when it is displayed on a computer monitor, but it is also often used to refer to the cell in a stored digital image that we cannot see until it is displayed. In raw spectral digital imagery each cell has a digital number (DN) or spectral reflectance value ranging from 0 to 255. These are grouped into clusters or categories during image classification that can be interpreted to represent features on the surface of the earth.

There are many methods of multispectral classification, but generally they fall into three groups:

- supervised classification
- unsupervised classification
- combinations of supervised and unsupervised classification

GRASS 4.0 supports supervised and unsupervised classification. Both types of classification require a two-step procedure. The two GRASS programs that create a supervised classified image are:

- i.class
- i.maxlik

Both programs must be run to complete a supervised classification.

The two programs that create an unsupervised image are:

- i.cluster
- i.maxlik

As with a supervised classification, both programs must be run to complete an unsupervised classification.

Other GRASS programs which can be used in the image classification process are:

- i.cca
- i.pca
- i.zc
- r.mapcalc

For more information on a specific command, select a command and press ESC.


GRASS Help

GRASS Documentation

GRASS WWW Home


Please use your browser buttons to navigate this documentation!
Automatically created on: Thu Jan 15 17:47:15 2004