一种基于模糊C均值聚类的多阈值苹果病害图像分割方法  

A multi-threshold segmentation method based on fuzzy C-means clustering for the apple disease leaf images

作　　者：贾庆节[1] 齐国红[1] 忽晓伟[1] JIA Qingjie;QI Guohong;HU Xiaowei(SIAS International University, Zhengzhou University, Xinzheng Zhengzhou 451150, China)

机构地区：[1]郑州大学西亚斯国际学院,河南新郑451150

出　　处：《智能计算机与应用》2019年第2期63-66,72,共5页Intelligent Computer and Applications

基　　金：河南省科技厅项目(182102210546);河南省高等学校重点科研项目基础研究计划(16A510034;17A520017);河南省教育厅2016年度民办高校品牌专业(自动化)建设项目(教政法[2016]896号);河南省教育厅第九批河南省重点学科(检测技术与自动化装置)建设项目(教高[2018]119号)

摘　　要：针对苹果病害叶片的图像分割问题,以斑点落叶病、花叶病和褐斑病为研究对象,提出一种基于模糊C均值聚类的多阈值分割算法。首先,将苹果病害叶片图像进行滤波;然后,利用模糊C均值进行病斑图像的聚类,再滤除病斑图像中的非病斑像素,根据分类结果获得分割阈值;最后,利用多阈值算法对苹果病害叶片图像进行分割,得到病斑图像。通过与其他分割方法进行比较表明,本方法分割出来的苹果病斑,分割准确率达到94%以上,分割效果明显。According to the properties of apple disease,using apple leaf ’s spot disease,mosaic disease and leaf spot as the research object,the paper proposes a multi-threshold segmentation algorithm based on Fuzzy C-means( FCM) clustering. Firstly,preprocess the image by filter to reduce the effect of noise;secondly,use FCM clustering method to divide the disease spot,and then eliminate some fake disease spot,achieve the thresholds according to the segmentation of the results;finally,use multi-threshold segmentation algorithm to segment the apple disease spot. Three kinds of apple disease leaf images are segmented effectively using the algorithm. The results of experiment indicate that apple disease spots can be separated precisely from the apple leaf images,the correct extraction rate can reach 94%. The research shows that the method is more valid than other method shown in this paper.

关 键 词：模糊C 均值聚类 苹果病害叶片 多阈值 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]