Abstract: There are still some problems on image matching for slow speed and poor robustness.The paper suggests a GGA approach to image matching，which is based on grey relational theory and genetic algorithm.In the method，matching parameter space is determined and several positions are acquired by the initialization of the chromosomes first.Secondly，a referential sequence and a comparative sequence are separately constructed by the histogram information of the template image and the current searching subimage.And then，taking the grey relational degree between the two sequences as the fitness function，most of the chromosomes concurrently approach to the matching position via basic principles of natural evolution，selection，crossover and mutation.The experimental results indicate that the algorithm not only obtains precise positions，but also obviously increases the matching speed and the ability to resist noise and geometrical distortion.

Key words: image matching, genetic algorithm, fitness function, grey relational theory

摘要： 针对图像匹配速度慢，抗干扰能力差的问题，将灰色关联理论与遗传算法相结合，提出了一种鲁棒性强的快速图像匹配方法——GGA（Grey Genetic Algorithm）法。该方法首先确定问题的参数空间，通过对参数空间编码和种群初始化得到待匹配的多个初始位置，然后利用模板图和当前搜索子图的直方图信息，分别构建参考序列和比较序列，并以两序列间的灰色关联度为适应度函数。在此基础上，初始群体经过选择、交叉和变异等操作逐代进化到搜索空间的优化区域，并逼近最佳匹配位置。实验结果显示，GGA法充分利用了灰色关联理论的小样本特性和遗传算法的计算并行性，在保证一定匹配精度的情况下，实时性和鲁棒性明显提高。

关键词: 图像匹配, 遗传算法, 适应度函数, 灰色关联理论