改进遗传算法在图像挖掘中的应用

Research of image mining based on improved genetic algorithm

图像关联规则挖掘是图像挖掘中重要研究课题。为了克服传统算法找到的关联规则数目太多,用户无法对其进行分析;种群进化陷入停滞不前,造成局部收敛等问题,引入了兴趣度这个阈值,提出了一种根据适应度值自动调整交叉概率和变异概率的新的自适应遗传算法,并成功地运用到Landsat卫星ETM数据图像,实现了图像关联规则的提取。实验结果表明,该算法在收敛快速性和稳定性等方面都有了明显改善,达到了预期效果。

Discovering image association rules is one of the most important tasks in image data mining.However,the number of discovered rules is so large,so the user cannot analyze all discovered rules.At the same time,the populations are troubled into stagnation,resulting in partial convergence.To solve those problems,the interest measure and a new adaptive genetic algorithm that bases the fitness value to adjust the crossover probability and mutation probability are concerned in the paper.Lastly,the algorithm successfully applied to mine the image association rules from Landsat satellite images.Some experiments show that the proposed new algorithm is clearly improved in convergent speed and stability and gets the expectation effect.