Shape matching for classification of historical glass

1
Shape matching for classification of historical
glass

• Laurens van der Maaten
• IKAT-Promovendidag 05

2
Introduction

• Classification of archaeological artifacts
• Now performed manually by experts
• Expert compares artifact with objects from
  reference collection
• Reference collections consist of drawings in
  books
• Thus slow, subjective, and error-prone process

3
Example

• An archaeological artifact

4
Example

• And its corresponding drawing
• ?

5
The task

• Given an artifact photograph
• Find the most alike drawings
• Speeds up the classification process
• Can give archaeological experts new insights

6
The problem

• Drawings contain no color information
• Drawings contain only very abstract texture
  representation
• Texture hard to extract from glass photographs
• Thus only outer shape information useful
• Shape matching

7
Shape matching

• Several approaches
• Shape contexts (Belongie, 2000)
• Curvature scale spaces (Mokhtarian, 1996)
• Turning functions (Veltkamp, 2003)
• Dynamic programming (Petrakis, 2002)
• Moment invariants (Hu, 1962)
• Hausdorff, Procrustes, etc.
• MPEG-7 standard CSS

8
Shape matching

• We compared various approaches
• Shape contexts
• Curvature scale space
• Desired properties of approach
• Invariant to scale, translation, and rotation
• Robust to distortions due to
• Broken artifacts
• 3D rotations
• Drawing interpretations

9
Broken artifacts

• ?

10
Or worse
11
3D rotations

• ?

12
Shape contexts

• Sample points from outer contour
• For all points
• Determine angle (relative to baseline) and
  distance to all other points in log-polar space
• All resulting histograms form the shape
  description

13
Shape contexts

• Matching using startpoint invariant k-NN
  classifier (using ?2-distance)
• Startpoint invariance obtained by circular
  shifting one of the histograms

14
Curvature scale space

• Determine positions of zero-crossings of
  curvature for an evolving shape contour
• Curvature is a function that is 1 for a straight
  line, and 1 / r for a circle with radius r
• Shape evolution convolve coordinates with 1D
  Gaussian kernel with increasing variance

15
Shape evolution

• Evolving shape with curvature zero-crossings

16
Curvature scale space

• CSS image
• Align CSS by aligning global maximum
• Sum distances between main peaks

17
Results

• Low identification performance (expected) due to
  difficult dataset

18
Results

• We examined various variations, such as
  quantization in shape context space, etc.
• Best performance 33 for hitlist size 10
• However
• For good artifacts results are encouraging
• Shape analysis on reference collection allows for
  making shape maps (using MDS)
• This allows for archaeologists to create new
  typologies (since archaeological typologies are
  not fixed)

19
Example query
?
20
(No Transcript)
21
Conclusions

• Matching glass artifacts with drawings is a
  difficult problem
• Shape context matching outperforms (MPEG-7
  standard) CSS matching
• Allows for shape analysis of reference collection
• Future research should focus extracting texture
  from photographs (e.g., Gabor)

22
Questions

• ?