Object Recognition

1
Object Recognition

• Tom McGrath
• CIS 601

2
What is object recognition?

• Perception of objects is different for humans
  than for computers.
• For humans perception of familiar items.
• For computers perception of familiar patterns.
• Are they really the same thing?

3
What do we mean by objects

• What we call object recognition may also be
  called pattern recognition.
• A pattern is an arrangement of descriptors.
• Descriptors may have more forms, but they are
  primarily vectors and strings.

4
More generally

• Object recognition is the process whereby
  observers are able to recognize three-dimensional
  objects despite receiving only two-dimensional
  input that varies greatly depending on viewing
  conditions.(2)

5
2 main approaches

• Decision-theoretic
• Patterns described using quantitative
  descriptors.
• Structural
• Patterns represented by symbolic information.
• Strings, for example.

6
Decision Theoretic

• Based on discriminant functions
• Let x (x1, x2, , Xn)T represent an
  n-dimensional pattern vector
• Let W (w1, w2, ,wW) be pattern classes.

7
Basic problem of decision-theoretic

• We want to find W decision functions d1(x),
  d2(x),,dw(x) with the property
• If a pattern x belongs to class wi, then
  di(x) gt dj(x),
• where j 1, 2, , W j ! i

8
In other words..

• We want to classify x, which is a pattern.
• We are given a finite set of classes of objects.
• We want to categorize the pattern x into one of
  the classes.
• To do so, we apply x to all decision functions,
  and categorize x to the class of best fit.

9
Structural

• Represents objects as strings, trees, graphs..
• Define descriptors and recognition rules base on
  the representations.

10
What does finite classification imply?

• The idea of a finite set of classes is quite
  limiting.
• Corresponds with industries use of object
  recognition very application specific.
• Indicates that computer object recognition
  techniques lack some abilities which are simple
  for humans.

11
Differences in classification

• Techniques thus far only classify objects based
  on their shape, color, texture, etc. These are
  only representative of the light reflected by an
  object.
• Humans classify objects many ways, including an
  objects function.

12
For example

• We classify a ring of rocks with a fire inside as
  a fire pit.
• We classify a board as a joist once it is
  installed as support for the floor.
• We classify our computer as a paperweight once it
  is more than five years old.

13
Correlation

• Given an image, we want to find all places in the
  image which contain a subimage, also called a
  template.
• Very useful for answering where is the x in
  this picture?

14
Notice..

• Recognition models typically rely on input from
  optical sensors.
• Such input is represented entirely in
  two-dimensional space.

15
Is a 3D representation necessary?

• DARPA challenge was not successfully completed.
• Armys LADAR sensors, which provide depth data,
  have demonstrated more capability.

16
3D Object recognition with neural trees

• First stage extracts features from the input
  range images.
• These features are used in the second stage to
  group image pixels into different surface patches
  according to the six surface classes proposed by
  the differential geometry.(4)

17
Invariants

• Basic idea
• D(g(A),g(B)) D(A,B)
• For all g in transformation group G
• Limitations
• There are very many possible transformations in
  G, and computation times becomes a problem.

18
Varying goals of object recognition

• Are we looking for that object?
• Face recognition
• Are we looking for one of those objects?
• Web search for 1987 Chevy pickup.

19
Notice

• Just because an object exists in an image doesnt
  mean it is recognizable.
• Example from Late Night with Conan OBrien

20
We dont know what this is
21
Recognizable as a human face
22
Recognizable as the pope
23
The Punchline
24
Histogram approach

• Vary bad results for images with
• Much noise
• Small target objects
• With tightly controlled conditions, moderate
  success can be achieved.

25
Noisy histograms
26
Noisy histograms
27
Noisy histograms
28
Correlation example

• Find the flower

29
Create template

• Actual template size 32X32

30
Acquire input image

• Actual image size 1600X1200

31
Compute correlation image

• Actual image size 1600X1200

32
Show areas of best match

• Actual image size 1600X1200

33
Find flower with more noise

• Source image 1600X1200

34
Correlation image
35
Area of best match?
36
Templates for a coin

• Acquire a template

37
Acquire target image

• Actual size 1600X1200

38
Compute correlation image
39
Display area of best match
40
Finding coin among noise
41
Correlation image
42
Brightest coin wrong one
43
Among different noise
44
Correlation image
45
Coin found
46
Structural approach to stapler

• Acquire source stapler image

47
Segment and find the stapler edge
48
Compute the boundary

• Image recreated from computed boundary

49
Select boundary points

• Boundary points at distance of 8

50
Image recreated from boundary points
51
Compare to a different view
52
Segment and acquire boundary
53
Image redrawn from boundary data
54
Boundary points selected at a distance of 8
55
Redrawn from selected boundary points
56
Final step

• Compare the chain code strings of the 2 sets of
  boundary points.

57
Finding boundaries with noise

• Custom filters for each target image may be
  required

58
Conclusions

• Modern object recognition techniques can provide
  much functionality in controlled environments.
• Simulation of human object recognition
  capabilities is a long way off.

59
Best way to search for objects

• The best approach to create an image search
  engine requires extensive human labor involving
  organizing every image in the database into its
  correct hierarchical position.
• Input as text can provide as much functionality
  as input from images in this approach.

60
References

• Digital Image Processing using Matlab
• Prentice Hall, ISBN 0-13-008519-7
• Michael Tarr, Brown University
• http//www.cog.brown.edu/tarr/pdf/Tarr02ECS.pdf
  search'object20recognition
• 3D Object Recognition by Neural
  Trees http//csdl.computer.org/comp/proceedings/i
  cip/1997/8183/03/81830408abs.htm
• Longjin Jan Latecki, CIS 601 Lecture notes
• http//www.cis.temple.edu/latecki/CIS601-04/lect
  ures_fall04.htm