当前位置: 首页 > 文章 > 基于K_means硬聚类算法的葡萄病害彩色图像分割方法 农业工程学报 2010,26 (2) 32-37
Position: Home > Articles > Segmentation of color images of grape diseases using K_means clustering algorithm Transactions of the Chinese Society of Agricultural Engineering 2010,26 (2) 32-37

基于K_means硬聚类算法的葡萄病害彩色图像分割方法

作  者:
李冠林;马占鸿;黄冲;迟永伟;王海光
单  位:
中国农业大学农学与生物技术学院永宁县设施园艺研究所;中国农业大学农学与生物技术学院;中国农业大学植物病理学系
关键词:
彩色图像分割;葡萄病害;颜色聚类;L*a*b*颜色空间;K_means硬聚类;相似度
摘  要:
为了提高植物病害图像的分割精度与效果,根据植物病害症状及图像的特点,提出了一种基于K_means硬聚类算法(HCM)的葡萄病害彩色图像非监督性分割处理方法。该方法是在L*a*b*颜色空间模式下利用ab二维数据空间的颜色差异,以平方欧式距离作为像素间的相似度距离、以均方差作为聚类准则函数对颜色进行二分类聚类,并通过数学形态学运算对聚类结果进行校正。利用该方法对3种葡萄病害彩色图像进行分割的结果表明,该方法能够较为准确地将病斑区域从彩色图像中分割出来,对葡萄病害彩色图像的分割处理比较理想,鲁棒性好,分割准确率高。
译  名:
Segmentation of color images of grape diseases using K_means clustering algorithm
作  者:
Li Guanlin1,Ma Zhanhong1,Huang Chong1,Chi Yongwei2,Wang Haiguang1(1.College of Agriculture and Biotechnology,China Agriculture University,Beijing 100193,China;2.Yongning Protected Horticulture Research Institute,Yongning 750100,China)
关键词:
color image segmentation,grape diseases,color cluster,L*a*b* color space,K_means clustering algorithms,similarity
摘  要:
To improve the segmentation precision and effectiveness of plant disease images,a kind of unsupervised segmentation processing method based on K_means clustering(HCM) algorithm was proposed according to the properties of the symptoms and images of plant diseases.On the basis of the color differences of ab two-dimension data space from L*a*b* color space model,iterative color clustering of two clusters was conducted using squared Euclidian distance as the similarity distance and mean square deviation as the clustering criterion function.And the mathematics morphology algorithm was used to correct the clustering results.The proposed method was used to segment the color images of three kinds of grape diseases.The results show that it can satisfactorily segment the diseased regions from the color images of grape diseases with good robustness and good accuracy.

相似文章

计量
文章访问数: 6
HTML全文浏览量: 0
PDF下载量: 0

所属期刊

推荐期刊