Vision & RecognitionVision & Recognition
• From a different direction
• At different times, particularly if it has been modified in the interval
• In different light
A particular object may have different visual attributes when viewed:
ObservationsObservations
• Be able to distinguish between different objects with the same appearance
• Be able to recognize the same object viewed from different directions or in different light
• Be able to recognize the same object after modification
• Be able to distinguish between an object and its reflection
ObjectivesObjectives
•Pattern recognition
•Robot vision
•Remotely operated devices
•Security related software
•Disease Diagnosis
ApplicationsApplications
5
The Minimum 3-D Feature Set The Minimum 3-D Feature Set
Two cubes of the same color are indistinguishable.
Cubes are distinguished by color, color count and color position
6
The Minimum 3-D Feature Set The Minimum 3-D Feature Set
2 sides color ‘A’
2 sides color ‘B’
Maximum of 2 distinguishable cubes
Cubes are distinguished by color, color count and color position
7
The Minimum 3-D Feature SetThe Minimum 3-D Feature Set
3 sides color ‘A’
3 sides color ‘B’
Maximum of 2 distinguishable cubes.
2 colors -> 8 cubesCubes are distinguished by color, color count and color position
8
The Minimum 3-D Feature SetThe Minimum 3-D Feature Set
Like color sides opposite vs. like colored sides adjacent.
3 colors -> 27 cubes
Cubes are distinguished by color, color count and color position
9
With 4 color cubes, pattern With 4 color cubes, pattern breaks. breaks.
We would expect 4We would expect 433 = 64 = 64
But we only get 42 But we only get 42 distinguishable cubes.distinguishable cubes.
Why? ---> Rotations !Why? ---> Rotations !
The Minimum 3-D Feature SetThe Minimum 3-D Feature Set
10
Rotations and ReflectionsRotations and Reflections
The Eight Queens Problem
for (i = 0; i < 8; i++){
ok = 1;
pNextPair->column = i;
pPairOnTop = pTop;
while (ok = 1 && pPairOnTop != NULL){
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Hidden patterns revealed by tiling solutions
Q Q Q Q QQ Q Q Q Q
Q Q Q Q QQ Q Q Q Q
Q Q Q Q QQ Q Q Q Q
Q Q Q Q QQ Q Q Q Q
Q Q Q Q QQ Q Q Q Q
Q Q Q Q QQ Q Q Q Q
Q Q Q Q QQ Q Q Q Q
Q Q Q Q QQ Q Q Q Q
Q Q Q Q QQ Q Q Q Q
Q Q Q Q QQ Q Q Q Q
Q Q Q Q QQ Q Q Q Q
Q Q Q Q QQ Q Q Q Q
Horizontal Shifts
One left succeeds, all others fail to find solution
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Vertical
Shifts
One up or one down succeeds; all others fail to find a solution
Q QQ Q
Q QQ Q
Q QQ Q
Q QQ QQ Q
Q QQ Q
Q QQ Q
Q QQ Q
Q Q
Q Q QQ Q Q
Q Q QQ Q Q
Q Q QQ Q Q
Q Q QQ Q Q
Rotation & Mirror Effects
Same object viewed from different angles
Q QQ Q
Q QQ Q
Q QQ Q
Q QQ Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Diagonal shifts& Mirror Effects
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q QQ Q Q Q
Q Q Q Q
Q QQ Q
Q QQ Q
Q QQ Q
Q QQ Q
Q QQ Q
Q QQ Q
Q QQ Q
Q QQ Q
Q QQ Q
Q QQ Q
Q QQ Q
Q QQ Q
Q QQ Q
Q QQ Q
Q QQ Q
Q QQ Q
Additional Mirror Effects
Q Q Q Q QQ Q Q Q Q
Q Q Q Q Q QQ Q Q Q Q Q
Q Q Q Q QQ Q Q Q Q
Q Q Q Q QQ Q Q Q Q Q
Q Q Q Q QQ Q Q Q Q
Q Q Q Q Q QQ Q Q Q Q Q
Q Q Q Q QQ Q Q Q Q
Q Q Q Q QQ Q Q Q Q Q
Q Q Q Q QQ Q Q Q Q
Q Q Q Q Q QQ Q Q Q Q Q
Q Q Q Q QQ Q Q Q Q
Q Q Q Q QQ Q Q Q Q Q
Q Q Q Q QQ Q Q Q Q
Q Q Q Q Q QQ Q Q Q Q Q
Q Q Q Q QQ Q Q Q Q
Q Q Q Q QQ Q Q Q Q QHow do we determine which patterns are
significant?
12 partitions
12 distinct solutions
Why not 96 (= 12 x 8) permutations?
The answer is that one of our partitions contains only 4 "aliases" because the solution is diagonally symmetric
Permutations of the solution Permutations of the solution vectorsvectors
Shape detection Shape detection algorithmsalgorithms
Without color…
example universe:
assembly-line where all parts are uniformly steel-colored
Shape detection Shape detection algorithmsalgorithms
Edge detection
Using changes in reflectivity
Shape detection Shape detection algorithmsalgorithms
Edge detection
Edges can intersect with planes, or air
Is the background the same object or a different one?
Shape data structShape data structuresures
Corners have “more information” per point
Collection of points and relative distances
Collection of lines and intersections
Collections of shape primitives
Shape data structShape data structuresures
Corners can be used to define
Collections of edges and intersections
Or boundaries of planes
Shape data structShape data structuresures
Skeletons…AKA
Wire-Frames
Advantage… easier to morph
Shape data structShape data structuresures
Volume filling shape primitives
advantage: can easily calculate volume as sum of primitives’ volume
There are many applications...There are many applications...
Volumes of research…Volumes of research…
And many areas yet And many areas yet unexplored...unexplored...
Vision & RecognitionVision & Recognition
Thank you.Thank you.
Top Related