Prematurity and slow convergence are two difficult problems in genetic algorithm, a new crossover is proposed, named hybrid crossover operator based on pattern, to gain a better convergence to the optimal solution. To retain the diversity of population, the approach of pattern is used, together with the behavior of antitone. The new method can be used for those application problems which are wanted to reach their best value quickly. More experiments show that the new crossover can find the global optimal solution and improve convergence ratio obviously.