Enhancement in performance of genetic algorithm for object location problem

The object location problem has been solved using the repeated genetic algorithm by determining the number of independent runs to guarantee a given probability of success. However, this number is still too large for the detection of some noisy images with acceptable certainty. Through an in depth analysis of all the genetic operations and their interrelationships, we design an improved crossover and a dynamic search scheme that integrate the crossover, mutation and selection operations so that the probability of success of correct location in a single run for some test objects is enhanced significantly. As a consequence, only a few repeated runs are required to guarantee a high probability of success in solving this type of real problem.