基于Fisher准则的免疫思维进化算法在图像阈值寻优中的应用  

An Immune Mind Evolutionary Algorithm with Fisher Criteria and Its Application in Image Segmentation

作　　者：张捷[1] 谢刚[1] 

机构地区：[1]太原理工大学信息工程学院,山西太原030024

出　　处：《太原理工大学学报》2009年第3期225-228,共4页Journal of Taiyuan University of Technology

基　　金：山西省青年学术带头人项目(晋教科函[2007]11号);教育部博士点基金(2006112005);山西省青年科学基金项目(2007021018)

摘　　要：针对图像分割中的阈值选取问题,通过引入模式识别理论中的Fisher评价函数作为图像分割的准则函数,提出了基于Fisher准则的免疫思维进化算法进行图像分割。该算法克服了基本遗传算法无法记忆每一代的进化信息、容易出现"早熟"、收敛慢且不能保证全局搜索的缺点,可以指导算法向着有利方向进行,并且采用群体搜索,提高了搜索效率且避免了"早熟"现象,提高了算法的寻优质量。笔者对该算法进行了详尽的分割实验,并与基本遗传算法相比较,结果表明该算法可以将分割阈值稳定在5个像素之内,运算速度比基本遗传算法提高了9.5%,是一种有效的图像阈值分割新方法。Aiming at the threshold selection problems in image segmentation, this paper proposed an immune mind evolutionary algorithm with Fisher criteria function for image segmentation by introducing the concept of Fisher criteria function in pattern recognition theory as the criteria function for image segmentation. The algorithm overcomes the shortcomings of basic genetic algorithm, such as no memory of the evolution information of each generation, premature, slow convergence, and no guarantee for global searching. By using global searching, this algorithm proceeds cowards a favorable direction, thus increasing searching efficiency, preventing ＂premature＂ phenomenon,and improving optimization quality. The comparison of the segmentation experiments by this algorithm and basic genetic algorithm indicates that this algorithm has a threshold stabilized within 5 pixels,with a computation speed 9.5% higher than genetic algorithm. There- fore,it is a new effective method for image segmentation.

关 键 词：免疫思维进化算法 Fisher准则函数 图像分割 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]