SBSM对开放环境下多色叶片RGB图像背景的精确分离  

Accurate Segregation for RGB Image Background of Multi-color Leaves by SBSM(Saturation Background Segregate Method)in an Open Environment

作　　者：赵羡波 林国健[2] 林志华[2] 曾文龙[2] 胡正平 包可翔 陈义强 张佩[4] ZHAO Xianbo;LIN Guojian;LIN Zhihua;ZENG Wenlong;HU Zhengping;BAO Kexiang;CHEN Yiqiang;ZHANG Pei(China Tobacco Fujian Industrial Co.,Ltd.,Xiamen,Fujian 361000;Longyan Company,Fujian Tobacco Corporation,Longyan,Fujian 364000;Zhangzhou Company,Fujian Tobacco Corporation,Zhangzhou,Fujian 363000;Jiangsu Meteorological Bureau,Nanjing,Jiangsu 210008,China)

机构地区：[1]福建中烟工业有限责任公司,福建厦门361000 [2]福建省烟草公司龙岩市公司,福建龙岩364000 [3]福建省烟草公司漳州市公司,福建漳州363000 [4]江苏省气象局,江苏南京210008

出　　处：《贵州农业科学》2020年第11期166-172,共7页Guizhou Agricultural Sciences

基　　金：福建省烟草公司龙岩市公司科技与创新项目“烟叶初分等级快速诊断系统的研发”(2020Y07);江苏省“333工程”培养资金资助项目(BRA2019348)。

摘　　要：为实验室及户外等开放环境下精确分离背景与目标叶片提供科学依据,选择9种类型多色叶片为研究对象,研究边缘检测法、通道差值法、颜色阈值法、最大类间方差法(OTSU)和图像饱和度背景分离法(SBSM)对开放环境下多色叶片RGB图像背景的分离效果。结果表明:在开放环境下,SBSM对多色叶片图像背景分离处理的时长适中,为6.70 s;对背景噪音、叶片阴影去除充分,对单色叶、杂色叶、多色叶的处理效果均最佳;对各种叶片分离的精确度除3号样品为84.07%外,其余样品均达96%以上,其对多色叶片图像背景分离的精准度最优,较边缘检测法、通道差值法、颜色阈值法和OTSU法分别提高72.19%、32.14%、27.46%和18.92%。The effects of RGB image background of nine types of multi-color leaves segregated by edge detection method,channel difference method,color threshold method,the maximum between-class variance method(OTSU)method and Saturation background segregate method(SBSM)in open environment are compared to provide the scientific basis for accurate segregation background and target leaves in an open environment such as laboratory and outdoors.Result:In an open environment,SBSM segregated multi-color leaves image background has the moderate duration with 6.70 s,fully removes background noise and leaf shadow,and has the best processing effect on mono-color leaves,variegated leaves and multi-color leaves.The segregation accuracy of all kinds of leaves is above 96%,except for No.3 samples of 84.07%.The accuracy of SBSM segregated multi-color leaves image background is the best,which is 72.19%,32.14%,27.46%and 18.92%higher than that of edge method,channel difference method,color threshold method and OTSU method.

关 键 词：植物叶片 多色叶 RGB图像 开放环境 背景分离 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]