基于高空间分辨率遥感数据的天山云杉树冠信息提取研究  

Study of crown information extraction of Picea schrenkiana var. tianschanica based on high-resolution satellite remote sensing data

作　　者：刘玉锋 潘英 李虎 LIU Yufeng;PAN Ying;LI Hu(College of Computer and Information Engineering,Chuzhou University,Chuzhou 239000,China;R&D Center of Data Products and Application Software on Anhui High Resolution Earth Observation System,Chuzhou 239000,China;Students Affairs Department,Chuzhou University,Chuzhou 239000,China)

机构地区：[1]滁州学院计算机与信息工程学院,滁州239000 [2]安徽省高分辨率对地观测系统数据产品与应用软件研发中心,滁州239000 [3]滁州学院学生处,滁州239000

出　　处：《国土资源遥感》2019年第4期112-119,共8页Remote Sensing for Land & Resources

基　　金：安徽省高等学校自然科学研究项目“复杂地形条件下高分卫星数据林木冠幅遥感估算”(编号:KJ2016A526);滁州学院校级培育项目“天山云杉林木冠幅高分遥感估算关键技术研究”(编号:2015PY04)共同资助

摘　　要：针对天山云杉在高空间分辨率卫星遥感图像上的成像呈圆形或近圆形特征,从空间几何特征入手进行森林立木树冠信息提取。将多尺度斑点检测和梯度矢量流(gradient vector flow,GVF)Snake主动轮廓模型有序地结合在一起,提出了“树冠顶点探测——树冠轮廓绘制——树冠轮廓优化”的树冠信息提取的技术流程,解决了标记分水岭变换目标标记难以准确设定、主动轮廓模型演化结果受制于轮廓线初始位置的问题,从而得到了定位准确、边界简洁的树冠轮廓结果。经与调查样地当中每一株树的实测冠幅值进行比较,该方法对高、中、低郁闭度的天山云杉树冠信息都有较好的提取结果,平均误差分别为10.8%,4.5%和6.4%。表明该方法适用于天山云杉林木树冠信息提取,在中亚山地森林资源遥感监测领域具有应用推广价值。For the reasons that images of Picea schrenkiana var.tianschanica in the western tianshan forest were round or suborbicular,crown information extraction was conducted with the space geometric features.According to the workflow of“investigating the features of Picea schrenkiana var.tianschanica in satellite images-extracting tree crown information-estimating tree crown width with remote sensing images”,a method for estimating tree crown width in Central Asia mountain forests based on remote sensing images was proposed and evaluated,with the purpose of challenging the problems that it is difficult to set the marks for marking watershed transform target ground objects and the active contour model evolution results are limited by the original positions of contour lines.Multi-scale blob detection,marking watershed transform and GVF Snake active contour model were orderly combined for tree crown information extraction.This technical process integrated and optimized the process of tree crown information extraction,and gained the tree crown contour distribution maPof Picea schrenkiana var.tianschanica from images.A comparison with the measured tree crown width of each tree in the investigated sample ground shows that this method well estimates the tree crown width of Picea schrenkiana var.Tianschanica with high,medium or low canopy density,with the mean error being 10.8%,4.5%and 6.4%,respectively.The research results provide a better solution for the key technical problem of tree crown interpretation for high-resolution remote sensing data in forest resource monitoring.

关 键 词：高空间分辨率遥感 天山云杉树冠 多尺度斑点检测 标记分水岭变换 GVF SNAKE模型 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]