Exam III

Question 1

What is a Kappa statistic?

Answer 1

A measure of reproducibility for categorical data

Question 2

2 methods of unsupervised image classification?

Answer 2

1. ) Histogram Based

2. ) Sequential Clustering

Question 3

What is Sequential clustering?

Answer 3

Assign pixels to closest class mean

Question 4

Define unsupervised classification.

Answer 4

• > Process of assigning pixels to classes
• > Grouping pixels based on similar DN’s
• > For remote areas
• > No prior knowledge
• > No human error
• > Unique/distinct classes

Question 5

How does histogram based U.C. work?

Answer 5

• > Bell shape
• > Peak = class
• > Limit = half the distance

Question 6

Steps in sequential clustering?

Answer 6

1. ) Pick # of classes
2. ) Pick radius for a spectral class
3. ) Based off min. distance to closest mean
4. ) Group two closest if one is exceeded

Question 7

Define supervised classification.

Answer 7

Grouping pixels based on statistics from selected training sites

Question 8

Advantages of supervised classification?

Answer 8

• > For known areas
• > Specific Areas of known identity
• > Can detect errors ahead of time

Question 9

Disadvantages of supervised classification?

Answer 9

• > Impose a classification on data

- > Training sites may not represent “everything”

Question 10

List of procedures for supervised classification?

Answer 10

1. ) Select training sites
2. ) Get statistics (X bar, med, mode, range)
3. ) Evaluate sig.s
4. ) Select algorithm
5. ) Classify area
6. ) Perform A.A. (Accuracy Assesment)

Question 11

Classification choices for supervised classification?

Answer 11

1. ) Parallelepiped - only bounds used, not stable
2. ) Min. distance to mean - doesn’t use “variability”
3. ) Maximum likelihood - considers X bar, std dev, variance

Question 12

Difference between precision, accuracy, bias?

Answer 12

Precision - degree of agreement
Accuracy - closeness to true value
Bias - direction

Question 13

Define accuracy assessment.

Answer 13

-> Comparing the map product with a reference of in situ (real world) measurements

Question 14

Sources of error in A.A.?

Answer 14

• > Similar spectral sig.s (H20, shadows)
• > Dissimilar spectral sig.s.
• > Mixed pixels**
• > Registration accuracy

Question 15

Non-site specific vs site specific A.A.?

Answer 15

• > Non-site compares based on area per class

- > Site specific compares locations between map and reference data at specific sites (error matrix)