SHAME: Shape Matching Environment (UIF.5055)

There are numerous applications where a geometric pattern of points, curves, and regions need to be checked and compared. An important class of applications are industrial inspection systems, which perform quality control in a production environment, often by camera inspection. Concrete examples are checking the correct positioning of labels on boxes and bottles, and verifying printings of wallpaper and textiles. Another class of applications are retrieval systems, which search an object from a large database. Concrete examples are searching for drilling holes in a collection of technical engineering drawings, and multimedia applications like the retrieval of a particular sculpture from a museum collection. Other types of applications include object recognition and classification, fingerprint matching, and stereo image matching.
In all these applications, a geometric pattern that consists of a set of points, curves, or regions, needs to be compared to another pattern. This can be done by transforming one pattern so as to align it as good as possible to another pattern, which is called 'shape matching'. There are numerous techniques for this stemming from image processing or statistical analysis, but all current techniques are limited in scope and applicability. In particular in those matching problems where shape (rather than intensity or colour) plays a crucial role, few useful techniques are known. In this project we use relatively new techniques from the field of computational geometry, which are much more suitable for matching based on shape. The state-of-the-art results in computational geometry on shape matching are all very recent. Unfortunately, many useful shape matching algorithms have not found their way into practice yet, because these algorithms are inherently complex. Also, most theoretical papers assume arithmetic over real numbers and ignore degenerate cases, while in practice often floating point arithmetic is used, and input data is often degenerate by construction. At the same time, we see that most of the problems solved are still relatively simple (matching of point sets, matching of two curves or two regions) while practical applications are more complex (point sets with accuracy weights, collections of curves and regions, articulated curves, shape deformation). Therefore, the objectives of this project are to further develop the theory in these more complex areas, and to develop a state of the art software shape matching environment (SHAME) of currently available algorithms.

These objective will be achieved in the following way:

• A scientific programmer will build a whole library of matching algorithms and a large flagship example, and will perform extensive experimental evaluation on large data sets. A PhD student will extend existing results for practical problems, including matching of weighted point sets, where the weights represent a measuring accuracy, and matching of articulated figures, like a pair of scissors.
• A post-doc will work on more complex problems, including matching of shape approximations and discrete contours, and the deformation of one shape into another.

The following companies have shown explicit interest in this project, and want to take a seat in the utilization committee:

• Algorithmic Solutions
  The objective of this company is to commercialize scientific, object oriented software libraries. The SHAME library fits well in their product line, and they are interested in exploring the possibility to commercialize it.
• Dutch Vision Systems (DVS)
  This company develops industrial inspection systems. Each new application requires its own custom-designed software, which can be partly based on the SHAME library.
• Ellips
  Ellips is specialized in designing vision systems for agricultural inspection. They develop application specific software for quality control (including shape) of fruit and vegetables.
• Philips Research
  The Graphics unit of Philips Research is interested in this project for use in 31) depth reconstruction and watermarking of images.
• TNO Technisch Physische Dienst (TNO-TPD)
  The shape matching functionality that comes available through the SHAME project provides complementary capabilities for their applications in pattern recognition.

Er zijn nog geen resultaten bekend.

Gebruikers