Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
12 Cards in this Set
- Front
- Back
Supervised Classification
|
User creates regions of interest (such as water, urban, vegetation). Computer uses spectral signatures of selected training data to categorize areas that do not fall within training data.
|
|
Supervised Classification: Advantages
|
User has control
Processing is linked to known locations (less confounding data) User doesn’t have to match up classification categories with field data Easier to detect and solve errors |
|
Supervised Classification: Disadvantages
|
User has to place some semblance of structure over the area which may lead to oversimplification
Training classes are usually based on field identification rather than spectral properties Training data selected by user may not be representative of the entire image as categories on a map are commonly heterogenous |
|
Unsupervised Classification
|
Automatic categorization by program based on spectral signature of each pixel. Computer groups pixels of similar spectral characteristics into unique clusters.
*Assumes that spectral classes within a given cover class should cluster together spatially (nearest neighbor) |
|
Unsupervised Classification: Advantages
|
Prior knowledge of the area being studied isn’t necessary
Minimizes human error as fewer decisions are made by the user Generally produces more uniform results Spectral distinction is harder for the human eye to discern over a computer Faster |
|
Unsupervised Classification: Disadvantages
|
Spectral groupings may not be of interest to the user
Spectral properties of classes change over time as the relationship between information classes and spectral classes is not constant Limited control over the types of classes classified by computer |
|
Maximum Likelihood Classification
|
Quantitatively evaluates variance and covariance of trained spectral patterns when evaluating an unclassified pixel
Classifier assumes the distribution of points for each cover type is normally distributed Distribution of each category response can be described by its mean vector and covariance matrix With these two values, the classifier computes the probability that an unclassified pixel will belong to a particular category/spectral class |
|
Object-Oriented Classification
|
Involves identification of image objects or segments that are spatially contiguous pixels of similar texture, color, and tone.
Allows for consideration of shape, size, and context as well as spectral content Often more effective than pixel-based methods when classifying high-res imagery |
|
Supervised Image Classification Procedure
|
Image > Supervised Training > Pixel Labeling > Accuracy Assessment
|
|
Unsupervised Image Classification Procedure
|
Image > Clustering & Cluster Analysis > Clustering & Cluster Grouping > Accuracy Assessment
|
|
Producer Accuracy
|
The probability that a certain land cover of an area on the ground is classified as such. This is the ratio of correctly classified pixels for a particular classification by total pixels OBSERVED as that same particular classification
|
|
User Accuracy
|
The probability that a pixel labeled as a certain land cover class is actually the specified class. This is the ratio of correctly classified pixels for a particular classification by total pixels CLASSIFIED as that same particular classification
|