基于超像素聚类的EM脐橙病虫害图像的分割  

Segmentation of EM Navel Orange Diseases and Pests Images Based on Superpixel Clustering

作　　者：刘永平 章银娥[1] LIU Yongping;ZHANG Yine(School of Mathematics and Computer Science,Gannan Normal University,Ganzhou 341000,China)

机构地区：[1]赣南师范大学数学与计算机科学学院,江西赣州341000

出　　处：《赣南师范大学学报》2021年第3期86-91,共6页Journal of Gannan Normal University

基　　金：国家自然科学基金项目(62041210);江西省教育厅教育科技项目(GJJ170817);教育部产学研协同项目(202002050031);江西省数值模拟与仿真技术重点实验项目(461055)。

摘　　要：随着赣南脐橙果业优良品质的发展,赣南脐橙越来越受到人们的青睐.但是在发展赣南脐橙精准果业的同时,脐橙病虫害也随之蔓延.因此快速分割和识别赣南脐橙病虫害的病斑图像,以便提高赣南脐橙的果业质量.本文提出一种基于混合聚类的脐橙病虫害图像分割方法,通过超像素聚类将整个彩色赣南脐橙病虫害图像分割成若干个紧凑的、近似均匀的超像素块,为图像分割提供有用的聚类线索,以加快期望最大化(EM)算法的收敛速度.使用EM算法从每个超像素块中快速准确地分割出病变像素,根据数学形态学去校正聚类结果,最终获得分割后的病斑图像.对比本实验和与同类方法的实验,实验结果表明该方法是有效的,对脐橙病虫害病斑图像的分割具有较好的实验效果.With the development of the excellent quality of the Gannan navel orange fruit industry,the Gannan navel orange has become more and more popular.However,while developing the precision fruit industry for navel oranges in southern Jiangxi,the pests and diseases of navel oranges also spread.Therefore,the diseased image of Gannan navel orange diseases and insect pests are segmented and identified quickly in order to improve the fruit industry quality of Gannan navel orange.A segmentation method based on hybrid clustering is proposed for navel orange pests and diseases in this paper.Through superpixel clustering,the entire color Gannan navel orange pest image is segmented into several compact and approximately uniform superpixel blocks,which provide useful clustering for image segmentation,to speed up the convergence speed of the Expectation Maximization(EM)algorithm.Lesion pixels are segmented quickly and accurately from each super-pixel block by using of The EM algorithm.The clustering results are corrected according to the mathematical morphology to obtain the segmented lesion image finally.Comparing this experiment with the experiment with similar methods,the results show that this method is effective and has a good experimental effect on the segmentation of the images of navel orange diseases and pests.

关 键 词：病虫害分割 超像素聚类 EM算法 

分 类 号：S123[农业科学—农业基础科学]