【作者】 张一栋；

【导师】 吴锡生；

【作者基本信息】 江南大学 ， 计算机应用技术， 2008， 硕士

【副题名】一种基于免疫遗传算法和粗糙集的改进图像分割算法

【摘要】 图像工程是近几年发展起来的一门学科,它的研究内容非常丰富,根据抽象程度和研究方法的不同可分为三个层次:图像处理、图像分析和图像理解。图像处理的目的之一是图像识别,而图像分割与测量是图像识别工作的基础。图像分割是将图像分成一些有意义的区域,然后对这些区域进行描述,相当于提取出某些目标区域图像的特征,判断图像中是否有感兴趣的目标。有学者提出了一种免疫网络算法在数据分析中的新设计思想,近来又有学者结合粗糙集提出了一种新的分割算法,该算法在人脑MRI图像分割实验中得到了很好的分割效果,但由于FCM算法本身的缺陷(对区域间的连通性不确定),在具体应用中总会出现一些问题。论文基于免疫遗传分类算法提出了一种通过无教师学习方法确定聚类数和聚类中心的免疫遗传聚类算法;接着,在一些学者研究工作的基础上,针对粗糙集中属性约简算法复杂度高的问题,提出了一种改进算法,降低了属性约简算法的复杂度;最后把无教师学习的免疫遗传算法和粗糙集结合得到了一种改进的分割算法,从而可以很好的避免前述算法需要事先设置聚类数和聚类中心、对区域的连通性不确定的缺陷,提高分割准确度,而且免疫遗传算法本身具有聚类速度快,精确度高的优点。经实验验证,提高了聚类的速度,对图像的分割更细致、更准确,提高了分割算法的效率。更多还原

【Abstract】 Image engineering is a subject that has been developed in recent years, and it has many contents. According to the degree of abstract and the investigate methods, the research on it can be divided into three levels: image processing, image analysis and image comprehension. One of image processing goals is the pattern recognition, but the image segmentation and the survey are the pattern recognition work foundations. The image segmentation is that the image is divided into to the significance some regions, then carried on the description to these regions, equal in withdrawing the characteristic of certain target sector images, at last judged image whether it has the interested goal.Some scholars proposed one new kind design concept of immunity network algorithm in the data analysis, recently also other scholars had gather the rough collection to propose one new kind of segmentation algorithm, this algorithm obtaining the very good segmentation effect in the human brain MRI image segmentation experiment, but because of FCM algorithm itself flaw (to connectivity between region indefinite), the algorithm has some problems in the general application.The paper based on the immunity heredity classification algorithm proposed one kind studies the method determination cluster number and the cluster center immunity heredity cluster algorithm through the non-teacher; Then, at in some scholar research work foundation, in view of the rough centralism attribute reduction algorithm order of complexity high question, proposed one kind of improvement algorithm, reducing the attribute reduction algorithm order of complexity; Finally the immunity genetic algorithm which the non-teacher studies builds up roughly gathers obtained one kind of improvement segmentation algorithm, avoids the forecited algorithm needing to establish the cluster number and the cluster center, to the region connective indefinite flaw beforehand, and enhances the segmentation accuracy.Moreover immunity genetic algorithm itself has the cluster speed to be quick and high precision merit.Confirmed after the experiment, the new algorithm enhances the cluster speed, is more careful to the image segmentation, accurately, and increases the segmentation algorithm efficiency.更多还原