基于微分进化的最大类间方差图像分割算法及应用  

The Differential Evolution based on the Otsu's Algorithm and Application of the Image Segmentation

作　　者：陈科尹[1] 邹湘军[1] 陈丽娟[1] 

机构地区：[1]华南农业大学南方农业机械与装备关键技术教育部重点实验室,广东广州510642

出　　处：《装备制造技术》2012年第2期47-49,共3页Equipment Manufacturing Technology

基　　金：教育部博士点资助项目(200805640009);广东省自然科学基金资助项目(9251064201000009;9151064201000030);广东省产学研企业科技特派员建设项目(2010B090500017);广东省部产学研结合科技创新平台(2011A091000017)

摘　　要：通过对传统的最大类间方差胶囊图像分割法的研究,提出了一种基于微分进化算法的最大类间方差的图像分割方法,并且成功于应用于基于机器视觉的胶囊检测对象中。该方法利用胶囊图像分割阈值的实数编码序列作为样本个体,选择最大类间方差作为个体的生存适应度评价标准,进行一系列微分进化算法运算后,最终获得胶囊检测图像的理想分割阈值。实验分析结果表明该方法不仅可以达到最大类间方差法的图像分割效果,而且运行效率优于最大类间方差的图像分割方法,可以提高机器视觉中的图像分割质量。Through studying the traditional Otsu segmentation method of the capsule image;We proposed a Otsu＇s algorithm based on the differential evolution.And this method was successfully applicated into the capsule detection based on the machine vision.The method uses the real number coding sequences of the threshold of the capsule image segmentation as the samples of the individuals;And chooses the Otsu＇s survival as an individual fitness for the evaluation criteria;Then computes a series of the differential evolution algorithm and finally obtains the ideal segmentation threshold of the capsule image for the capsule detection.The experimental results show that this method can not only achieve the maximum between-class variance method of the image segmentation;And is better than the Otsu＇s segmentation method in the efficiency,improve the quality of the image segmentation.

关 键 词：图像分割法 最大类间方差法 微分进化算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]