Abstract: This paper improves the genetic algorithm for labeling line drawings presented by Richard Myers in the following areas: self-adaption parameter adjustment method was used, the hybridization rate and mutation rate of individuals with high fitness in the same generation were dynamically changed, and the rate of hybridization and mutation of individuals with lower fitness was set as a fixed value; the constraints were established when the initial population was created, with the aim of improving the uncertainty of the initial population coverage space and the relative irrationality of the individual distribution; the fitness function of the genetic algorithm was modified so that the selection operator with the individual fitness as the index could correctly guide the algorithm to search the solution space. The genetic algorithm was employed to mark six different line drawings, the parameter a and c in the formula of hybridization rate and mutation rate were used as experimental variables, the change tendency of the algorithm’s mark success rate curve was analyzed, and the influence of operator parameter setting on the performance of genetic algorithm was discussed. The results show that c belongs to interval [0, 0.05] and a belongs to interval [0.8, 1.0] and a is the optimal parameter setting of the genetic algorithm for labeling line drawings.

Key words:  line drawing label, genetic algorithm, binary coding, fitness