Scribd
Upload
3 views

Lecture7a Classification

Uploaded by

f5dry54d5b
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
3 views

Lecture7a Classification

Uploaded by

f5dry54d5b
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 1

ENVS406 Remote Sensing (Fall 2023)

Lecture Outline 10-10-2023


Classification of Multispectral Data

General Goals/Steps of classification:


1. Identify what it is you want to identify/classify
Examples: forest vs nonforest, snow cover, wet vs dry?
2. Decide how to distinguish one feature (or class of object) from another*
Examples: do you want to decide what feature in an image is what, or do you want to
use some kind unbiased, statistical method?
3. Develop a template/reference of the representative spectral characteristics (signatures) for
each feature (or class)
4. Apply the template/reference to the original image and cross-walk the spectral
characteristics in every pixel in the image to what it has been classified as
5. Evaluate your results
6. Refine your reference/template if desired (step 3), and repeat step 4-5

*Unsupervised classification
 Often the first step in any classification, and sometimes used to generate some of the
information required for the supervised classification below.
 Doesn’t require foreknowledge of classes or their spectral signatures
 Groups pixels into x desired number of classes based on the similarity of spectral signatures
 Common methods include “K-means”, “isocluster”, and “migrating means clustering”

*Supervised classification
 In most cases, this is ultimately the method of choice to extract information from an image
 Does require foreknowledge of the classes and requires a training sample, which defines the
spectral signatures for each feature/class
 Assigns pixels to classes by “matching” the spectral signatures in the training sample to the
pixel spectral signatures
 Common method is “maximum likelihood classification” (MLC), rooted in basic statistics

In practice, both unsupervised and supervised classification are compared and often combined

You might also like