摘要
叶脉网络的提取及其性状参数的测算,可为植物叶脉生态学机理研究提供重要参考。以不同叶特性的6类树种(国槐、毛白杨、臭椿、洋白蜡、元宝枫和栾树)叶片为对象,基于e Cognition软件对叶脉显微图像进行多尺度分割,综合利用显微图像的光谱信息和几何信息构建提取知识库,并使用叶脉循环生长法对提取结果进行完善,增加叶脉网络的完整性。结果表明,叶脉提取的最优阈值分别为:尺度参数200,形状参数0. 7,紧凑度参数0. 3,亮度特征值230~280,光谱特征值180~230,几何特征值大于1. 5。叶脉密度测算的精度均达到了93%以上,对植物叶脉信息的快速提取具有较高的普适性。
The extraction of leaf network and the measurement of its trait parameters provide an important reference for the study of leaf vein ecology. Taking the leaves of six tree species ( Sophora japonica, Populus tomentosa, Ailanthus altissima, Fraxinus pennsylvanica, Acer truncutum and Koelreuteria paniculata ) with different leaf characteristics as object, the multi-scale segmentation of the vein microscopy image was based on eCognition software. Firstly, the microscopic images were segmented. And then the spectral information and object geometry information of microscopic images objects were comprehensively applied to build the road extraction knowledge base. Thirdly, the results of vein extraction were improved and completed in order to increase the integrity of the vein network. The results showed that the optimal thresholds for leaf vein extraction were: scale parameter was 200, shape parameter was 0.7, tightness parameter was 0.3, brightness characteristics value was 230~280, spectral characteristic value was 180~230, geometric feature value was greater than 1.5. The extraction of leaf vein density measurement was more than 93%, which had high universality.
作者
朱济友
于强
YANG Di
徐程扬
岳阳
陈向
ZHU Jiyou;YU Qiang;YANG Di;XU Chengyang;YUE Yang;CHEN Xiang(College of Forestry, Beijing Forestry University, Beijing 100083, China;Department of Geography, University of Florida, Gainesville FL 326113, USA;Guangzhou Urban Planning and Design Survey Research Institute, Guangzhou 510060, China)
出处
《农业机械学报》
EI
CAS
CSCD
北大核心
2019年第1期51-57,共7页
Transactions of the Chinese Society for Agricultural Machinery
基金
中央高校基本科研业务费专项资金项目(BLX201806)
林业公益性行业科研专项重大项目(20140430102)
中国博士后科学基金面上项目(2018M641218)
关键词
叶脉网络
叶脉密度
显微图像
面向对象法
leaf vein network
leaf vein density
microscopic image
object-oriented method