Inexact graph matching using a genetic algorithm for image recognition

Exact graph matching using a genetic algorithm for image recognition has been introduced in previously published work. The algorithm was based on angle matching between two given graphs. It has proven to be quite effective in exact graph matching. However, the algorithm needs some modifications in order to handle cases where the number of nodes, shapes and rotations of the two graphs are different. This paper presents modifications such as the introduction of node exemption, inexact matching between straight lines and curves in the graphs and consideration of rotational degrees of the graphs. Each angle in a graph is also given a weight to indicate the significant degree of identifying the graph. A multi-objective function is used to reflect the similarity between two graphs. The experiments designed to evaluate the algorithm have shown very promising results. It is highly accurate in matching graphs with dissimilarities in shape, number of nodes and degrees of rotation.