改进分水岭算法在无人机遥感影像树冠分割中的应用  

Forest canopy segmentation of UAV remote sensing images using improved watershed algorithm

作　　者：于旭宅 王瑞瑞[1,2] 陈伟杰 YU Xuzhai;WANG Ruirui;CHEN Weijie(Department of Forestry Beijing Forestry University,Beijing 100083,China;Beijing Key Laboratory of Precision Forestry,Beijing Forestry University,Beijing 100083,China;Tianjin Jinghai Highway Engineering Co.,Ltd.,Tianjin 301600,China)

机构地区：[1]北京林业大学林学院,北京100083 [2]北京林业大学精准林业北京市重点实验室,北京100083 [3]天津市静海公路工程有限公司,天津301600

出　　处：《福建农林大学学报（自然科学版）》2018年第4期428-434,共7页Journal of Fujian Agriculture and Forestry University:Natural Science Edition

基　　金：北京林业大学优秀青年教师科技支持专项计划(YX2014-09)

摘　　要：提出了一种基于NDVI植被指数计算的改进分水岭分割方法.利用该方法对原始无人机多光谱遥感影像进行波段甄选、NDVI指数计算、形态学滤波等预处理,得到树冠的显著性区域图像;再利用彩色向量空间梯度算法计算显著性区域图像的梯度,从显著性区域图像中提取树冠的顶点及其范围作为标记,加到梯度图像上;最后采用基于标记控制的分水岭算法对树冠层进行分割.结果表明,该算法能够有效去除输电线路等背景区域的影响,算法样本精度达到88.3%.An improved watershed segmentation method based on NDVI vegetation index is proposed. This method firstly preprocessed the band selection,NDVI index calculation and morphological filtering of the original UAV multispectral remote sensing image,and obtained the significant regional image of the canopy. Then the vertex and its range of the canopy are extracted from the salient regional images as markers and the tags are added to the gradient image. Finally,a watershed algorithm based on marker control is used to divide the canopy. The experimental results show that the algorithm can effectively remove the influence of the transmission line and other background regions. Compared with the results of the tree crowns extracted from visual interpretation,the accuracy of the sample algorithm was 88.3%.

关 键 词：输电线路通道 无人机多光谱影像 分水岭分割 植被指数 形态滤波 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]