1. Institute of Information Theory and Automation Introduction to Pattern Recognition Jan Flusser flusser@utia.cas.cz

2. Recognition (classification) = assigning a pattern/object to one of pre-defined classes Pattern Recognition

3. Recognition (classification) = assigning a pattern/object to one of pre-defined classes Statictical (feature-based) PR - the pattern is described by features (n-D vector in a metric space) Pattern Recognition

4. Recognition (classification) = assigning a pattern/object to one of pre-defined classes Syntactic (structural) PR - the pattern is described by its structure. Formal language theory (class = language, pattern = word) Pattern Recognition

5. Supervised PR – training set available for each class Pattern Recognition

6. Supervised PR – training set available for each class Unsupervised PR (clustering) – training set not available, No. of classes may not be known Pattern Recognition

7. PR system - Training stage Definition of the features Selection of the training set Computing the features of the training set Classification rule setup

8. Desirable properties of the features Invariance Discriminability Robustness Efficiency, independence, completeness

9. Desirable properties of the training set It should contain typical representatives of each class including intra-class variations Reliable and large enough Should be selected by domain experts

10. Classification rule setup Equivalent to a partitioning of the feature space Independent of the particular application

11. PR system – Recognition stage Image acquisition Preprocessing Object detection Computing of the features Classification Class label

12. An example – Fish classification

13. The features: Length, width, brightness

14. 2-D feature space

15. Empirical observation For a given training set, we can have several classifiers (several partitioning of the feature space)

16. Empirical observation For a given training set, we can have several classifiers (several partitioning of the feature space) The training samples are not always classified correctly

17. Empirical observation For a given training set, we can have several classifiers (several partitioning of the feature space) The training samples are not always classified correctly We should avoid overtraining of the classifier

18. Formal definition of the classifier Each class is characterized by its discriminant function g(x) Classification = maximization of g(x) Assign x to class i iff Discriminant functions defines decision boundaries in the feature space

19. Minimum distance (NN) classifier Discriminant function g(x) = 1/ dist(x,  ) Various definitions of dist(x  One-element training set 

20. Voronoi polygons

21. Minimum distance (NN) classifier Discriminant function g(x) = 1/ dist(x,  ) Various definitions of dist(x  One-element training set  Voronoi pol NN classifier may not be linear NN classifier is sensitive to outliers  k-NN classifier

22. Find nearest training points unless k samples belonging to one class is reached

23. Discriminant functions g(x) are hyperplanes Linear classifier

24. Assumption: feature values are random variables Statistic classifier, the decission is probabilistic It is based on the Bayes rule Bayesian classifier

25. The Bayes rule A posteriori probability Class-conditional probability A priori probability Total probability

26. Bayesian classifier Main idea: maximize posterior probability Since it is hard to do directly, we rather maximize In case of equal priors, we maximize only

27. Equivalent formulation in terms of discriminat functions

28. How to estimate ? Parametric estimate (assuming pdf is of known form, e.g. Gaussian) Non-parametric estimate (pdf is unknown or too complex) From the case studies performed before (OCR, speech recognition) From the occurence in the training set Assumption of equal priors

29. Parametric estimate of Gaussian

30. d-dimensional Gaussian pdf

31. The role of covariance matrix

32. Two-class Gaussian case in 2D Classification = comparison of two Gaussians

33. Two-class Gaussian case – Equal cov. mat. Linear decision boundary

34. Equal priors Classification by minimum Mahalanobis distance If the cov. mat. is diagonal with equal variances then we get “standard” minimum distance rule max min

35. Non-equal priors Linear decision boundary still preserved

36. General G case in 2D Decision boundary is a hyperquadric

37. General G case in 3D Decision boundary is a hyperquadric

38. More classes, Gaussian case in 2D

39. What to do if the classes are not normally distributed? Gaussian mixtures Parametric estimation of some other pdf Non-parametric estimation

40. Non-parametric estimation – Parzen window

41. The role of the window size Small window  overtaining Large window  data smoothing Continuous data  the size does not matter

42. n = 1 n = 10 n = 100 n = ∞

43. The role of the window size Small window Large window

44. Applications of Bayesian classifier in multispectral remote sensing Objects = pixels Features = pixel values in the spectral bands (from 4 to several hundreds) Training set – selected manually by means of thematic maps (GIS), and on-site observation Number of classes – typicaly from 2 to 16

45. Satellite MS image

46. Other classification methods in RS Context-based classifiers Shape and textural features Post-classification filtering Spectral pixel unmixing

47. Typically for “YES – NO” features Feature metric is not explicitely defined Decision trees Non-metric classifiers

48. General decision tree

49. Binary decision tree Any decision tree can be replaced by a binary tree

50. Real-valued features Node decisions are in form of inequalities Training = setting their parameters Simple inequalities  stepwise decision boundary

51. Stepwise decision boundary

52. Real-valued features The tree structure and the form of inequalities influence both performance and speed.

55. How to evaluate the performance of the classifiers? - evaluation on the training set (optimistic error estimate) - evaluation on the test set (pesimistic error estimate) Classification performance

56. How to increase the performance? - other features - more features (dangerous – curse of dimensionality!) - other (larger, better) traning sets - other parametric model - other classifier - combining different classifiers Classification performance

57. Combining (fusing) classifiers C feature vectors Bayes rule: max max Several possibilities how to do that

58. Product rule Assumption: Conditional independence of x j max

59. Max-max and max-min rules Assumption: Equal priors max (max ) max (min )

60. Majority vote The most straightforward fusion method Can be used for all types of classifiers

61. Training set is not available, No. of classes may not be a priori known Unsupervised Classification (Cluster analysis)

62. What are clusters? Intuitive meaning - compact, well-separated subsets Formal definition - any partition of the data into disjoint subsets

63. What are clusters?

64. How to compare different clusterings? Variance measure J should be minimized Drawback – only clusterings with the same N can be compared. Global minimum J = 0 is reached in the degenerated case.

65. Clustering techniques Iterative methods - typically if N is given Hierarchical methods - typically if N is unknown Other methods - sequential, graph-based, branch & bound, fuzzy, genetic, etc.

66. Sequential clustering N may be unknown Very fast but not very good Each point is considered only once Idea: a new point is either added to an existing cluster or it forms a new cluster. The decision is based on the user-defined distance threshold.

67. Sequential clustering Drawbacks: -Dependence on the distance threshold -Dependence on the order of data points

68. Iterative clustering methods N-means clustering Iterative minimization of J ISODATA Iterative Self-Organizing DATa Analysis

69. N-means clustering 1.Select N initial cluster centroids.

70. N-means clustering 2. Classify every point x according to minimum distance.

71. N-means clustering 3. Recalculate the cluster centroids.

72. N-means clustering 4. If the centroids did not change then STOP else GOTO 2.

73. N-means clustering Drawbacks - The result depends on the initialization. - J is not minimized - The results are sometimes “intuitively wrong”.

74. N-means clustering – An example Two features, four points, two clusters (N = 2) Different initializations  different clusterings

75. Initial centroids N-means clustering – An example

76. Initial centroids N-means clustering – An example

77. Iterative minimization of J 1.Let’s have an initial clustering (by N-means) 2.For every point x do the following: Move x from its current cluster to another cluster, such that the decrease of J is maximized. 3. If all data points do not move, then STOP.

78. Example of “wrong” result

79. ISODATA Iterative clustering, N may vary. Sophisticated method, a part of many statistical software systems. Postprocessing after each iteration -Clusters with few elements are cancelled -Clusters with big variance are divided -Other merging and splitting strategies can be implemented

80. Hierarchical clustering methods Agglomerative clustering Divisive clustering

81. Basic agglomerative clustering 1.Each point = one cluster

82. Basic agglomerative clustering 1.Each point = one cluster 2.Find two “nearest” or “most similar” clusters and merge them together

83. Basic agglomerative clustering 1.Each point = one cluster 2.Find two “nearest” or “most similar” clusters and merge them together 3.Repeat 2 until the stop constraint is reached

84. Basic agglomerative clustering Particular implementations of this method differ from each other by - The STOP constraints - The distance/similarity measures used

85. Simple between-cluster distance measures d(A,B) = d(m 1,m 2 ) d(A,B) = min d(a,b) d(A,B) = max d(a,b)

86. Other between-cluster distance measures d(A,B) = Hausdorf distance H(A,B) d(A,B) = J(AUB) – J(A,B)

87. Agglomerative clustering – representation by a clustering tree (dendrogram)

88. How many clusters are there? 2 or 4 ? Clustering is a very subjective task

89. How many clusters are there? Difficult to answer even for humans “Clustering tendency” Hierarchical methods – N can be estimated from the complete dendrogram The methods minimizing a cost function – N can be estimated from “knees” in J-N graph

90. Life time of the clusters Optimal number of clusters = 4

91. Optimal number of clusters

92. Applications of clustering in image proc. Segmentation – clustering in color space Preliminary classification of multispectral images Clustering in parametric space – RANSAC, image registration and matching Numerous applications are outside image processing area

93. References Duda, Hart, Stork: Pattern Clasification, 2 nd ed., Wiley Interscience, 2001 Theodoridis, Koutrombas: Pattern Recognition, 2 nd ed., Elsevier, 2003 http://www.utia.cas.cz/user_data/zitova/prednasky/