A Genetic Algorithm for Labeling Point Features
Günther Robert Raidl · 2003
This paper introduces a genetic algorithm (GA) for tagging point features on images with text labels. The goal is to place all labels in a way that minimizes overlaps and simultaneously consider predefined position preferences. The proposed GA includes several problem dependent improvements: First, a preprocessing step reduces the search space in a safe way. Second, the starting population of the GA is generated in a heuristic way, which enables a faster convergence but nevertheless ensures the presence of enough diversity. Third, each newly generated solution is locally improved before its evaluation. The proposed GA is empirically compared to a very efficient simulated annealing approach using several randomly generated test cases.