摘要
在RGB和HSV空间中做图像分割时存在一些问题,比如分割精度低、处理进度慢等.为了解决这些问题,在K均值聚类分割方法的基础上,通过L*a*b*颜色空间进行分割的方法被提出.首先将原始RGB空间利用XYZ空间转换到L*a*b*颜色空间中,在L*a*b*颜色空间模式中使用a*b*二维数据空间的色差,K均值聚类算法的参数被不断调整,并通过数学形态学去校正聚类结果.最后得到病斑图像.本文通过该方法对4种脐橙病虫害进行分割,实验的结果表明,病斑区域能够在本文提出的方法中较为准确地把分割出来,并且对脐橙4种病虫害彩色图像的分割效果理想,显著提高了准确率,同时也表现出了该方法的竞争性.
There are some problems in image segmentation in RGB and HSV space,such as low segmentation accuracy and slow processing progress.In order to solve these problems,on the basis of k-means clustering segmentation method,the method of L*a*b*color space segmentation is proposed.Firstly,the original RGB space is transformed into L*a*b*color space by XYZ space.In the mode of L*a*b*color space,the chromatic aberration of a*b*two-dimensional data space is used.The parameters of k-means clustering algorithm are constantly adjusted,and the clustering results are corrected by mathematical morphology.Finally,images of disease spots were obtained.In this paper,four kinds of navel orange diseases and pests were segmented by this method.The experimental results showed that the disease spot area could be accurately segmented by the method proposed in this paper,and the segmentation effect of color images of four kinds of navel orange diseases and pests was ideal,which significantly improved the accuracy,and showed the competitiveness of this method.
作者
张红霞
章银娥
ZHANG Hongxia;ZHANG Yine(School of Mathematics and Computer Science,Gannan Normal University,Ganzhou 341000,China)
出处
《赣南师范大学学报》
2019年第6期44-48,共5页
Journal of Gannan Normal University
基金
江西省教育厅科技项目(GJJ1708072)
江西省教育厅人文社科规划项目(JC1610243)
江西省科技厅科技支撑计划项目(20151BBF60071)
江西省教育厅研究生教学改革项目(JXYJG-2018-159)
赣南师范大学校级研究生教改项目(200069)。
关键词
颜色空间
K均值聚类
形态学运算
图像分割
color space
K-means clustering
morphological operations
image segmentation
