Scribd
Upload
5 views8 pages

Object Recognition-12

Object Recognition chapter-12 from book of digital image processing by Gonzalez

Uploaded by

recidol890
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
5 views8 pages

Object Recognition-12

Object Recognition chapter-12 from book of digital image processing by Gonzalez

Uploaded by

recidol890
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 8

Object Recognition

Neural Networks
What are Neural Networks?
• Modeled after the brain’s neural structure.
• Consist of artificial neurons (nodes) connected like brain neurons.
• Each neuron receives inputs, processes them, and produces an output.
Key Historical Milestones
• 1943 - McCulloch and Pitts Model: First concept of a neural network.
• 1949 - Hebb’s Rule: Connections (synapses) strengthen between neurons that fire
together, foundational for learning.
• 1958 - Perceptron by Rosenblatt: Simple neural model used as a basis for future
networks.
• 1986 - Backpropagation: Enabled more complex learning in multi-layer networks.
Why Use Neural Networks for Object Recognition?
• Can recognize complex patterns (e.g., identifying faces in images).
• Handle non-linear problems well, which is important for real-world images.
• Adaptable to new data and good for tasks that need constant learning.
Structural Pattern Recognition
• Purpose: Recognizes complex patterns by identifying basic sub-
patterns (primitives) and their relationships.
• Focuses on shapes and structures, unlike statistical approaches that
rely on numerical data.
• Useful in fields like handwriting recognition, speech analysis, and
biometrics.
Components:
• Primitives: Basic building blocks (e.g., lines, curves).
• Syntactic Grammars: Define rules for combining primitives.
• Relational Graphs: Represent and compare relationships among primitives.
Structural Pattern Recognition
• Purpose: Recognizes complex patterns by identifying basic sub-patterns
(primitives) and their relationships.
• Focuses on shapes and structures, unlike statistical approaches that rely on
numerical data.
• Useful in fields like handwriting recognition, speech analysis, and
biometrics.
Components:
• Primitives: Basic building blocks (e.g., lines, curves).
• Syntactic Grammars: Define rules for combining primitives.
• Relational Graphs: Represent and compare relationships among
primitives.
Structural Pattern Recognition
Statistical Approach
• Focus: Counts segments only, providing
quantitative data.
• Limitation: Lacks insight into segment
organization or connection.
Structural Approach
• Focus: Emphasizes relationships and
arrangement of segments.
• Advantage: Ideal for complex patterns,
providing meaningful structural insight.

Fig: Statistical and structural approach


Matching Shape Numbers
• Compares region boundaries using shape numbers.
• Degree of similarity k is the largest order where shape numbers
coincide.
• Distance (D) between two shapes is the inverse of similarity k:
Dab​=1/k​
• Larger k → more similar shapes.
String Matching
• Objects are represented by strings of symbols. Matching algorithms
compare these strings to identify similar objects.
• Boundaries coded into strings of symbols.
• Matches occur when symbols at the same position in both strings are
identical.
• Similarity measure RRR is the ratio of matching symbols:
R(a,b) = a/max(∣a∣,∣b∣)
• R: Infinite for perfect match, 0 for no match.
String Matching

Fig:(a) Shapes. (b) Similarity tree. (c) Similarity matrix

You might also like