基于遗传算法的图像匹配算法研究

Image Matching Algorithm Based on Genetic Algorithm

图像匹配算法的研究是模式识别学科研究的一个分支,针对图像匹配速度问题,论文提出一种改进的遗传算法的快速图像匹配算法。该算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素子图的直方图信息,分别构建参考序列和比较序列,以两序列间的灰色关联度为适应度函数。通过改进的遗传算子实现遗传操作,对遗传个体进行迭代寻优,找出图像中的最佳匹配点。实验结果表明,基于该算法的图像匹配具有运算量小、匹配精确等优点,实时性得到了明显的提高。

The study of image registration algorithm is one of the branches in the field of pattern recognition. Aim at the problems on image matching for slow speed, an image matching algorithm based on a new genetic algorithm is proposed. In the method, matching parame- ter space is determined and several positions are acqulred 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 ap- proach to the matching position throug the new genetic algorithm. The experimental results inicate that the algorithm not only obtains precise positions, but alse obviously increases the matching speed.