Modeling Shapes for Pattern Recognition: A Simple Low-Cost Spline-based Approach

Abstract

We present a simple procedure for modeling shapes and trajectories of points using cubic polynomial splines. The procedure may prove useful for researchers working in the field of pattern recognition that are in the search of a simple functional representation for shapes and which are not particularly interested in diving into the hightheoretical aspects of more complex representations. The use of splines brings in a few advantages with regards to data dimensionality, speed and accuracy of processing, with minimal effort required for the implementation part. We describe several algorithms for data reduction, spline creation and query for which we provide pseudo code procedures in order to demonstrate the ease of implementation. We equally provide measurements on the approximation error and rate of data reduction.

Keywords

• spline
• cubic polynomial
• pattern recognition
• algorithms
• shape modeling
• curves
• approximation error
• Haussdorff distance