An Improved Genetic Algorithm for Solving Conic Fitting Problems

This paper presents an improved genetic algorithm for solving conic fitting problem. We first use several parallel small-populations genetic algorithms to obtain initial population, which has better average fitness. The range of mutation operator is also set to be gradually reduced with the growing of generation to guarantee the proportion of outstanding individuals within the population. An experiment shows that our improvements on genetic algorithm can remarkably increase the average fitness of population during evolution and enhance the performance of the algorithm as a whole.