改进遗传算法在潜艇天文导航中的应用

Application of Improved Genetic Algorithm to Submarine Celestial Navigation

潜艇自动天文导航对星图处理有着较高的实时性要求,在比较、分析、综合现有的一些改进方法的基础上,对基本遗传算法提出了一系列的改进措施,形成了一种高效综合的遗传算法.改进算法采用了最优保存策略和比例选择相结合的选择思路,设计了与进化代数相关的交叉概率和与个体适应度相关的变异概率,以及与早熟情况、进化代数和个体适应度有关的移民算法等.将该算法与大津法相结合应用于星图分割中,结果表明,改进的遗传算法不仅可以使图像分割获得满意的效果,而且还有效地提高了计算速度.

The real time processing of star image is required for submarine celestial navigation. On the basis of comparing, analyzing and synthesizing some available modifications, this paper presents series of modifications to GA, and forms a highly effective synthetic GA. In this algorithm, elitist model and fitness proportional model are composed for selection, a new probability algorithm of crossover depending on the number of generations and a new probability algorithm of mutation depending on the fitness value are designed. Furthermore, an improved immigration method depending on the convergence, evolution generation and fitness values is also proposed. Finally, this algorithm combined with Otsu has been applied to star-fields segmentation. The experiment result shows that using this algorithm not only satisfied effect for image segmentation but also quicker compute speed will be obtained.