【作者】 王宗跃；

【导师】 黄樟灿；

【作者基本信息】 武汉理工大学 ， 计算机应用技术， 2006， 硕士

【摘要】 本文主要分为三部分,第一部分为研究背景的介绍,给出了遗传程序设计(Genetic Programming,简称GP)的发展过程,从非线性GP发展到线性GP,详细地介绍了GEP的生物学背景、方法、研究现状。第二部分为MEP算法的研究,着重介绍了(Multi-Expression Programming,MEP)的编码方式和基本流程,并对其进行了分析,在此基础上,提出了改进的MEP算法并对该算法进行了详细的数值试验。第三部分为MEP算法的应用实践,将所设计的算法应用到图像配准问题,并将结果与传统的方法进行比较,结果显示本文设计的算法表现出了良好的性能。本文的主要工作和创新点如下: 1、对MEP独特的编码方式和相应的遗传操作进行了研究,与其他GP相比,MEP具有包含多个表达式、编码利用率高、不包含无用编码、不需要转化为树结构、能更好地保护好子结构等优点。 2、对MEP的不足之处进行改进。针对MEP和其他GP都存在搜索空间太大,无法进行有效搜索的问题,本文提出了针对不同问题预设复合函数模板和分级策略,从而减少了无用的搜索空间,提高了搜索效率。并用一些测试例子对改进的MEP算法进行实验,实验结果表明,改进的MEP算法是有效的。 3、将改进的MEP算法运用到图像配准问题。本文首先对图像配准问题进行了一定的概括,给出了图像配准的意义、变化类型、基于控制点的配准步骤等。并分析了现有模型的优缺点。然后给出了改进的MEP算法运用到图像配准问题的详细步骤,并做了详细的实验,取得了较好的结果。这是到目前为止,第一次将MEP算法运用到图像配准问题上。 4、编制了改进的MEP算法的软件。给出了该软件的框架和关键算法的代码。更多还原

【Abstract】 This thesis mainly divides into three parts. The first part for the research background introduction produced the GP developing process: from non-linear GP (genetic programming) to linear GP and introduced the GEP (Gene Expression Programming) biology background, the method, the research present situation in detail. The second part for the MEP (Multi-Expression Programming) algorithm research, emphatically introduced the MEP encoding method and the basic flow, and carried on the analysis to it. In this foundation, we proposed the improvement MEP algorithm and did the detailed numerical experimentation to this algorithm. The third part for the MEP algorithm application practice, applied the proposed algorithm to the image registration, compared and the results between the new and the traditional method, finally demonstrated our algorithm displaying good performance. Our main work and innovation as follows:1. We have conducted the research to the MEP unique encoding method and the corresponding evolutionary operation. Compared with other GP, MEP has many merits, such as encompassing many expressions, effectively using codes, no useless codes, without needing to transform to the tree structure, and safeguarding well on good sub- structure.2. We have made the improvement to the MEP deficiency. For the search space in MEP and other GPs is too large and unable to carry on the valid search, this thesis proposed different composite function templates and graduation strategies in view of the different questions. Thus this way reduced the useless search space, enhanced the search efficiency. With some test examples to the improvement MEP algorithm, the experimental results indicated that the improvement MEP algorithm is valid.3. We have utilized the improved MEP algorithm to the image matching question. This thesis has first summarized the image matching question, the significance of image matching, the transformation type, the matching process based on the control point and so on. After analyzing advantages and disadvantages on the existing model, we gave the detailed process about the improvement MEP algorithm applying on the image matching question, did the experiment, andobtained the good result. This is so far the first time utilizing the MEP algorithm to the image matching question.4. We have realized the improvement MEP algorithm software by designing the software frame and the essential algorithm codes.更多还原