基于DOM及LiDAR的多尺度分割与面向对象林隙分类

发布时间：2018-04-02 12:01

本文选题：林隙　切入点：影像分割　出处：《农业机械学报》2017年09期

【摘要】：为研究分割尺度对航空正射影像(DOM)与LiDAR数据协同面向对象林隙分割与分类的影响,以东北典型的天然次生林帽儿山实验林场东林施业区为试验区,对DOM与LiDAR数据进行多尺度分割与面向对象林隙分类。分割过程中,采用基于DOM分割、基于LiDAR数据分割、DOMLiDAR协同分割3种分割方案。每种分割方案采用10种尺度。在每种尺度应用两种数据提取的光谱和高度两种特征,采用支持向量机分类器(SVM)进行林隙分类。研究结果表明:3种分割与分类方案分类精度随尺度的增大整体呈现下降的趋势,与ED3(Modified)趋势相反。基于LiDAR数据在尺度参数10获得了最优分割结果。在所有尺度上(10~100),基于LiDAR数据分割与分类精度高于其他两种数据源的分类精度,相比单独使用DOM优势更加明显。基于LiDAR数据分割与分类方案在尺度参数10时获得了最高分类精度(Kappa系数为80%)。3种分割与分类方案最优尺度的分类精度显著高于其他尺度分类精度。分割尺度对面向对象林隙分类结果有重要影响。
[Abstract]:In order to study the effect of segmentation scale on the segmentation and classification of spatial gap between DOM (aerial orthophoto image) and LiDAR data, the Donglin construction area of Maoershan Experimental Forest Farm, a natural secondary forest in northeast China, was used as the experimental area.Multi-scale segmentation and object-oriented gap classification for DOM and LiDAR data are carried out.In the process of segmentation, there are three segmentation schemes based on DOM segmentation and LiDAR data segmentation based on DOMLiDAR.Each scheme uses 10 scales.The spectral and height features of two kinds of data were used in each scale, and support vector machine classifier (SVM) was used for gap classification.The results show that the classification accuracy of the three kinds of segmentation and classification schemes decreases with the increase of scale, which is contrary to the trend of ED3Modified.The optimal segmentation results are obtained based on LiDAR data at scale parameter 10.On all scales, the accuracy of data segmentation and classification based on LiDAR is higher than that of the other two data sources, and the advantage of using DOM alone is more obvious.Based on the LiDAR data segmentation and classification scheme, the highest classification accuracy is obtained at 10:00 of the scale parameters. The classification accuracy of the optimal scale is significantly higher than that of other scales.Segmentation scale has an important influence on the result of object-oriented gap classification.
【作者单位】：东北林业大学林学院;东北农业大学水利与土木工程学院;农业部农业水资源高效利用重点实验室;
【基金】：国家自然科学基金项目(31300533) 农业部农业水资源高效利用重点实验室开放课题项目(2015001、2015003)
【分类号】：S771;TP751

【相似文献】