introduction of pattern recognition.pdf

7/29/2019 Introduction of Pattern Recognition.pdf

1/40

PATTERN

RECOGNITION

Team teaching1


2/40

OUTLINES

Whats is pattern? What is class pattern? What is pattern recognition? Human perception Application example Statistically way Human and machine perception Pattern recognition process Topic Searching

2


3/40

WHAT IS A PATTERN?

A pattern is an abstract object, or a set ofmeasurements describing a physical object.

3


4/40

WHAT IS A PATTERN CLASS?

A pattern class (or category) is a set ofpatterns sharing common attributes.

A collection of similar (not necessarilyidentical) objects.

During recognition given objects are assignedto prescribed classes.

4


5/40

WHAT IS PATTERN RECOGNITION?

Theory, Algorithms, Systems to put Patternsinto Categories

Relate Perceived Pattern to PreviouslyPerceived Patterns

Learn to distinguish patterns of interest fromtheir background

5


6/40

HUMAN PERCEPTION

Humans have developed highly sophisticatedskills for sensing their environment and taking

actions according to what they observe, e.g.,

Recognizing a face. Understanding spoken words. Reading handwriting. Distinguishing fresh food from its smell.

We would like to give similar capabilities tomachines.

6


7/40

EXAMPLES OF APPLICATIONS

Handwritten: sorting letters by postal code.Printed texts: reading machines for blindpeople, digitalization of text documents.

Optical CharacterRecognition

(OCR)

Face recognition, verification, retrieval.Finger prints recognition.Speech recognition.Biometrics

Medical diagnosis: X-Ray, EKG(ElectroCardioGraph) analysis.

Diagnostic

systems

Automated Target Recognition (ATR).Image segmentation and analysis (recognitionfrom aerial or satelite photographs).

Militaryapplications

7


8/40

PATTERN RECOGNITION APPLICATIONS

BY PROBLEM DOMAINS

8


9/40

PATTERN RECOGNITION MODEL

9


10/40

THESTATISTICAL

WAY10


11/40

GRID BY GRID COMPARISON

AA B

Grid by Grid

Comparison

11


12/40


AA B

12

0 0 1 00 0 1 00 1 1 11 0 0 11 0 0 1

0 1 1 00 1 1 00 1 1 01 0 0 11 0 0 1

No ofMismatch= 3


13/40


AA B

Grid by Grid

Comparison

13


14/40


AA B

14

0 0 1 00 0 1 00 1 1 11 0 0 11 0 0 1

1 1 1 00 1 0 10 1 1 10 1 0 11 1 1 0

No ofMismatch= 9


15/40

PROBLEM WITH GRID BY GRIDCOMPARISON

Time to recognize a pattern- Proportional tothe number of stored patterns ( Too costly

with the increase of number of patterns

stored )

15

SolutionArtificial

Intelligence

A-Z a-z 0-9

*/-+1@#


16/40

HUMAN AND MACHINE PERCEPTION

We are often influenced by the knowledge of howpatterns are modeled and recognized in nature when we

develop pattern recognition algorithms.

Research on machine perception also helps us gaindeeper understanding and appreciation for pattern

recognition systems in nature.

Yet, we also apply many techniques that are purelynumerical and do not have any correspondence innatural systems.

16


17/40

PATTERN RECOGNITION

Two Phase : Learningand Detection.

Time to learn is higher. Driving a car

Difficult to learn but once learnt it becomesnatural.

Can use AI learning methodologies such as: Neural Network. Machine Learning.

17


18/40

BASIC CONCEPT

- Cannot be directlymeasured.

- Patterns with equalhidden state belong to

the same class.

Feature vector-Patterns with equal

hidden state belong tothe same class.

Task- To design aclassifer (decision

rule) which decidesabout a hidden state

based on an

observation.

Hidden state Feature vector

18


19/40

EXAM

PLE

Task: jockey-hoopsterrecognition.

The set of hiddenstate Y is

{H,J}

The feature space is

X2

19


20/40

LEARNING

How can machine learn the rule from data?

Supervised learning: a teacher provides a category label orcost for each pattern in the training set.

Unsupervised learning: the system forms clusters or naturalgroupings of the input patterns.

20


21/40

Classification(known categories) Clustering(creation of new categories)

CLASSIFICATION VS. CLUSTERING

21

Category A

Category B

Clustering(Unsupervised Classification)

Classification(Supervised Classification)


22/40

PATTERN RECOGNITION PROCESS(CONT.)

22

Post- processing

Classification

FeatureExtraction

Segmentation

Sensing

input

Decision


23/40

PATTERN RECOGNITION PROCESS

Data acquisition and sensing: Measurements of physical variables. Important issues: bandwidth, resolution , etc.

Pre-processing: Removal of noise in data. Isolation of patterns of interest from the background.

Feature extraction: Finding a new representation in terms of features.

Classification Using features and learned models to assign a pattern

to a category.

Post-processing Evaluation of confidence in decisions. 23


24/40

24

Sistem PR

Sensors andpreprocessing

Feature

extractionClassifier

Class

assignment

Learning algorithmTeacher

Pattern


25/40

CASE STUDY

Fish Classification: Sea Bass / Salmon.

Problem: Sorting incoming fishon a conveyor belt according to

species.

Assume that we have only two kinds of fish: Sea bass. Salmon.

25

Salmon

Sea-bass


26/40

CASE STUDY (CONT.)

What can cause problems during sensing? Lighting conditions. Position of fish on the conveyor belt. Camera noise. etc

What are the steps in the process?1.Capture image.2.Isolate fish3.Take measurements4.Make decision

26


27/40

CASE STUDY (CONT.)

27

Classification

FeatureExtraction

Pre-processing

Sea Bass Salmon


28/40

CASE STUDY (CONT.)

Pre-Processing: Image enhancement Separating touching or occluding fish. Finding the boundary of the fish.

28


29/40

HOW TO SEPARATESEA BASS FROM SALMON?

Possible features to be used: Length Lightness Width Number and shape of fins Position of the mouth Etc

Assume a fisherman told us that a sea bass isgenerally longer than a salmon.

Even though sea bass is longer than salmon on theaverage, there are many examples of fish where this

observation does not hold.29


30/40

HOW TO SEPARATE

SEA BASS FROM SALMON?

To improve recognition, we might have to usemore than one feature at a time. Single features might not yield the best performance. Combinations of features might yield better performance.

30

1

2

x

x

1

2

:

:

x lightness

x width


31/40

FEATURE SELECTION

31

Good

features Bad features


32/40

DECISION BOUNDARY

32


33/40

DECISION BOUNDARY (CONT.)

33More complex model result more complex boundary


34/40


34Different criteria lead to different decision boundaries


35/40


What if a customers find Sea bass in thereSalmon can?

We should also consider costs of differenterrors we make in our decisions.

35


36/40


For example, if the fish packing companyknows that:

Customers who buy salmon will object vigorouslyif they see sea bass in their cans.

Customers who buy sea bass will not be unhappyif they occasionally see some expensive salmon in

their cans.

How does this knowledge affect our decision?

36


37/40

CASE STUDY (CONT.)

Issues with feature extraction: Correlated features do not necessary improve

performance.

It might be difficult to extract certain features. It might be computationally expensive to extract

many features.

Missing Features. Domain Knowledge.

37


38/40

THE DESIGN CYCLE

Collecting training and testing data.Collect Data

Domain dependence.

Chose Features.

Domain dependence.Chose Model

Supervised learningUnsupervised learning.

Train

Performance with future dataEvaluate

38


39/40

Q & A39


40/40

TOPIC

SEARCHING

40

introduction of pattern recognition.pdf

Documents