Setting the Parameters of the LFT Shape Matching Algorithm

Abstract

The LFT algorithm (Large Fat Tetrahedron) is used to detect congruent subsets amongst unordered point sets and forms the kernel of a partial shape matching method. Although the method yields several advantages and is relatively efficient, its performance depends highly on the choice of various geometrical threshold parameters, as e.g., for the length difference of two edges. We present an overview of the key parameters of the algorithm and their influence on the computation, a guide line to provide an initial value for the parameters and we propose an approach to their automatic adjustment.