基于数值遗传算法的快速模板匹配

Fast Template Matching Based on the Numerical Genetic Algorithm

针对目前图像处理中模板匹配方法一般具有较大计算量的不足 ,提出使用遗传算法进行快速的搜索。由于图像本身是离散的模型 ,因而提出使用双线性插值算法 ,将搜索空间扩张到一个平面上的连续域 ,从而可以进行数值优化 ,目标函数为模板和子图像间的互相关。采用基于空间划分的数值遗传算法 ,通过在凸集上的杂交机制生成迭代解 ,具有较快的计算速度和较强的全局寻优能力。通过实验分析了目标函数的性质 ,并在最后给出了求解实例来说明算法的性能。

Many template matching approaches in current image processing literatures are not practical because of their time consuming and complicated computation. A fast search mechanism by using genetic algorithm is proposed. As an image is an intrinsically discrete model, we apply the bilinear interpolation algorithm to transform the searched space into a plane with a continuous domain so as to process the numerical optimization. The objective function is inter correlative between template and subimage. We use the numerical genetic algorithm, and obtain recursive solutions by crossover on a convex set. An experiment is given to analyze the properties of the objective function, and an example is shown to explain the performance of the algorithm.