This paper presents an improved niche genetic algorithm applied to multimodal function optimization for finding all the extreme solutions. This algorithm is pre-selected niche based and similarity based on the mechanism of eliminating the niche combination. We improve the adaptive crossover operator and mutation operator according to the probability, crossover probability and mutation probability the fitness value to dynamically adjust the individual. And the algorithm is used to solve a typical multi peak, the experimental results show that the niche genetic algorithmcan searchall themultimodal functions' optimal solutions and extreme solutions, and has faster search speed. At the same time, this method is universal in the multi peak function for the other.