1. Image Classification
To automatically categorize all pixels in an image into land cover classes or themes

2. Image Classification
Usually, multispectral data are used for this purpose: the spectral pattern within the data for each pixel is used as the basis for categorization
Different feature types show different combinations of DNs based on their inherent spectral reflectance and emittance properties

3. Image Classification
Pattern: set of radiance measurements in the various wavelength bands for each pixel
Spectral pattern recognition:
automated land cover classification using this pixel-by-pixel spectral information

4. Image Classification Spatial pattern recognition:
Categorization of image pixels based on their spatial relationship with their surrounding pixels
Image texture, feature size, shape are some of the criteria used

5. Image Classification Temporal pattern recognition:
Using time as an aid in feature identification
Classification methods are combinations of the above modes
(spectral, spatial, and temporal)

6. Image Classification Why classify? Make sense of a landscape
Place landscape into categories (classes)
Forest, Agriculture, Water, etc
Classification scheme = structure of classes
Depends on needs of users

7. Example Uses Provide context Landscape planning or assessment
Research projects
Drive models
Global carbon budgets
Meteorology
Biodiversity

8. Example: Near Mary’s Peak
Derived from a 1988 Landsat TM image
Distinguish types of forest

9. Classification
Basic strategy for classifying remotely-sensed images using spectral information
Supervised Classification
Unsupervised Classification

10. Basic Strategy Use radiometric properties of remote sensor
Different objects have different spectral signatures

11. Basic Strategy
In an easy world, all “Vegetation” pixels would have exactly the same spectral signature
Then we could just say that any pixel in an image with that signature was vegetation
We would do the same for soil, etc. and end up with a map of classes

12. The same would happen for other types of pixels, as well.
Basic Strategy
But in reality, that isn’t the case. Looking at several pixels with vegetation, you would see variety in spectral signatures.
The same would happen for other types of pixels, as well.

13. The Classification: Deal with variability
Different ways of dealing with the variability lead to different ways of classifying images
To talk about this, we need to look at spectral signatures a little differently

14. Think of a pixel’s reflectance in 2-dimensional space
Think of a pixel’s reflectance in 2-dimensional space. The pixel occupies a point in that space.
The vegetation pixel and the soil pixels occupy different points in 2-d space

15. In a Landsat scene, instead of two dimensions, we have six spectral dimensions
Each pixel represents a point in 6-dimensional space
To be generic to any sensor, we say “n-dimensional” space

16. Feature space image
A graphical representation of the pixels by plotting 2 bands vs. each other
Band 3
Band 4

17. Basic Strategy: Dealing with variability
With variability, the vegetation pixels now occupy a region, not a point, of n-dimensional space
Soil pixels occupy a different region of n-dimensional space

18. Basic strategy: Dealing with variability
Classification:
Delineate boundaries of classes in n-dimensional space
Assign class names to pixels using those boundaries

19. Classification Three types of classification Supervised
Requires “training pixels”, pixels where both the spectral values and the class is known.
Unsupervised
No extraneous data is used: classes are determined purely on difference in spectral values.
Hybrid
Use unsupervised and supervised classification together

20. Supervised Classification
Supervised classification requires the analyst to select training areas where he/she knows what is on the ground and then digitize a polygon within that area
Mean Spectral Signatures
The computer then creates
Conifer
Known Conifer Area
Digital Image
Water
Known Water Area
Deciduous
Known Deciduous Area

21. Supervised Classification
Mean Spectral Signatures
Information
Multispectral Image
(Classified Image)
Conifer
Deciduous
Water
Unknown
Spectral Signature of Next Pixel to be Classified

22. The Result is Information--in this case a Land Cover map
Legend:
Water
Conifer
Deciduous

23. Supervised Classification
Common Classifiers:
Parallelpiped
Minimum distance to mean
Maximum likelihood

24. Parallelepiped
The minimum and maximum DNs for each class are determined and are used as thresholds for classifying the image.
Benefits: simple to train and use, makes few assumptions about character of the classes.
Drawbacks: pixels in the gaps between the parallelepipes can not be classified; pixels in the region of overlapping parallelepipes can not be classified.

25. Parallelepiped

26. Parallelepiped

27. Parallelepiped

28. Supervised Classification: Statistical Approaches
Minimum distance to mean
Find mean or average spectral value of pixels of training sets in n-dimensional space
An unknown pixels is classified by computing the distance between the value of unknown pixel and each of the category means

29. Minimum distance

30. Supervised Classification: Minimum Distance
Benefits:
All regions of n-dimensional space are classified
Allows for diagonal boundaries (and hence no overlap of classes)

31. Supervised Classification
Band 3
Band 4
Minimum distance
Limitations:
Assumes that spectral variability is same in all directions, which is not the case
It is insensitive to different degrees of variance in the spectral response data
It is not used in applications where spectral classes are close to one another in the measurement space and have high variance
For most pixels, Band 4 is much more variable than Band 3

32. Minimum distance

33. Supervised Classification: Maximum Likelihood
Assume that the distribution of the cloud of points forming the category training data is Gaussian (normal)
For each class, build a discriminant function
For each pixel in the image, this function calculates the probability that the pixel is a member of a particular land cover class
Max likelihood uses the variance and covariance in class spectra to determine classification scheme.
Each pixel is assigned to the class for which it has the highest probability of membership

35. Maximum likelihood- The probability values plotted in a three dimensional graph
The vertical axis is associated with the probability of a pixel value being a member of one of the classes.
The resulting bell-shaped surfaces are called probability density function

36. Maximum Likelihood Classifier
Mean Signature 1
Candidate Pixel
Relative Reflectance
Mean Signature 2
It appears that the candidate pixel is closest to Signature 1. However, when we consider the variance around the signatures…
Blue
Green
Red
Near-IR
Mid-IR

37. Maximum Likelihood Classifier
Mean Signature 1
Candidate Pixel
Relative Reflectance
Mean Signature 2
The candidate pixel clearly belongs to the signature 2 group.
Blue
Green
Red
Near-IR
Mid-IR

38. Maximum Likelihood Classifier

39. Supervised Classification
Maximum likelihood
Benefits:
Most sophisticated; achieves good separation of classes
Drawbacks:
Requires strong training set to accurately describe mean and covariance structure of classes

40. Supervised Classification
In addition to classified image, you can construct a “distance” image
For each pixel, calculate the distance between its position in n-dimensional space and the center of class in which it is placed
Regions poorly represented in the training dataset will likely be relatively far from class center points
May give an indication of how well your training set samples the landscape

41. Unsupervised Classification
Unsupervised classifiers do not utilize training data
They involve algorithms that examine the unknown pixels in an image and group them into a number of classes based on the clusters (spectral classes) present in the scene
The analyst then determines the land cover identity of the clusters by comparing the image data to ground reference data

42. Unsupervised Classification
The analyst requests the computer to examine the image and extract a number of spectrally distinct clusters…
Spectrally Distinct Clusters
Cluster 3
Cluster 5
Cluster 1
Cluster 6
Cluster 2
Cluster 4
Digital Image

43. Unsupervised Classification
Output Classified Image
Saved Clusters
Cluster 3
Cluster 5
Cluster 1
Cluster 6
Cluster 2
Cluster 4
Next Pixel to be Classified
Unknown

44. Unsupervised Classification
The result of the unsupervised classification is not yet information until…
The analyst determines the ground cover for each of the clusters
???
Water
???
Water
???
Conifer
???
Conifer
???
Hardwood
???
Hardwood

45. Unsupervised Classification
It is a simple process to regroup (recode) the clusters into meaningful information classes (the legend).
The result is essentially the same as that of the supervised classification:
Conif.
Hardw.
Water
Land Cover Map
Legend
Water
Conifer
Hardwood
Labels

46. Unsupervised Classification
Benefits
The classifier identifies the distinct spectral classes present in the image data
Drawbacks
The maximally-separable clusters in spectral space may not match our perception of the important classes on the landscape

47. Unsupervised Classification
Clustering algorithms that can be used to determine the natural spectral groupings present in a data set:
“K-means” approach accepts from the analyst the number of clusters to be located in the data and the algorithm then seeds arbitrarily the cluster centers.

48. Unsupervised Classification
Clustering algorithms (continued):
Pixels are then assigned to the cluster whose arbitrary mean vector is closest. Revised mean vectors are computed and data are reclassified until no significant change appears.

49. Unsupervised Classification

50. Unsupervised Classification
Algorithms that incorporate a sensitivity to image texture or roughness
Texture is defined by the multidimensional variance observed in a moving window passed through the image (e.g. 3x3)

51. Unsupervised Classification
Threshold of variance determines smooth (homogeneous) or rough (heterogeneous) windows
The mean of the first smooth window becomes the first cluster center

52. Hybrid Classification
Unsupervised training areas
Contrary to supervised classification, these are areas chosen to contain non-homogeneous cover types.
Valuable in cases of variations within cover types and/or due to site conditions.
Guided clustering

53. The Output Stage Graphic products Tabular data
Digital information files

54. Postclassification Smoothing
Classified output is smoothed to show only the dominant classification
Majority filter

55. Classification: Summary
Use spectral (radiometric) differences to distinguish objects
Land cover not necessarily equivalent to land use
Supervised classification
Training areas characterize spectral properties of classes
Assign other pixels to classes by matching with spectral properties of training sets
Unsupervised classification
Maximize separability of clusters
Assign class names to clusters after classification

56. Spectral Clusters and Spectral Signatures
Recall that clusters are spectrally distinct and signatures are informationally distinct
When using the supervised procedure, the analyst must ensure that the informationally distinct signatures are spectrally distinct
When using the unsupervised procedure, the analyst must supply the spectrally distinct clusters with information (label the clusters).

57. Spectrally Distinct Signatures
Most image processing software have a set of programs which allow you to:
Graphically view the spectral signatures
Compute a distance matrix (measuring the spectral distance between all pairs of signature means)
Analyze statistics and histograms etc...
After you analyze the signatures, the software should allow you to:
Modify merge or delete any signatures
Remember--they must be spectrally distinct!
Finally, you can then classify the imagery (using a maximum likelihood classifier).

58. Evaluating Signatures--Signature Plots

59. Evaluating Signatures--Signature Ellipses

60. Evaluating Signatures--Signature Ellipses