Discrete artificial electric field algorithm for high-order graph matching

Abstract

High-order graph matching is a problem of establishing the correspondences between two sets of visual features subject to high-order matching constraints. This is an NP-hard combinatorial optimization problem and formulated as a maximization problem of matching score over all permutations of features. Artificial electric field algorithm (AEFA) (Yadav et al., 2019) is a proven optimization algorithm in the family of meta-heuristic and performed well for continuous optimization problems. In this article, we extended the AEFA algorithm for combinatorial high-order graph matching problems and introduced a discrete artificial electric field algorithm (DAEFA). This framework incorporates the redefine position and velocity representation scheme, addition–subtraction operation, velocity and position update rules, and a problem specific initialization by using heuristic information. The efficiency of the proposed algorithm is tested over three well-known datasets: synthetic, CMU house and real-world datasets. The computational results measured the matching score, accuracy of matching and established the correspondences between two graphs. The computational results show the outperformance of the proposed algorithm over the other state-of-art algorithms in terms of good matching score and accuracy both.