文献类型：期刊文献

英文题名：Extraction of Information on Trees outside Forests Based on Very High Spatial Resolution Remote Sensing Images

作者：Sun, Bin[1,2] Gao, Zhihai[1] Zhao, Longcai[3] Wang, Hongyan[3] Gao, Wentao[3] Zhang, Yuanyuan[1]

第一作者：孙斌;Sun, Bin

通信作者：Wang, HY[1]

机构：[1]Chinese Acad Forestry, Inst Forest Resource Informat Tech, 1 Dongxiaofu, Beijing 100091, Peoples R China;[2]European Space Agcy, European Space Res Inst ESRIN, Via Galileo Galilei, I-00044 Frascati, Italy;[3]Chinese Acad Sci, Inst Remote Sensing & Digital Earth RADI, Beijing 100094, Peoples R China

年份：2019

卷号：10

期号：10

外文期刊名：FORESTS

收录：;EI(收录号：20194207561051);Scopus(收录号：2-s2.0-85073443073);WOS:【SCI-EXPANDED(收录号:WOS:000498395600012)】；

基金：This research was funded by the Fundamental Research Funds for the Central Non-profit Research Institution of CAF, grant number CAFYBB2019ZB004, National Science and Technology Major Project of China, grant number 21-Y20A06-9001-17/18, National Natural Science Foundation of China, grant number 41501467, and ESA-MOST China Dragon 4 Cooperation, grant number 32396.

语种：英文

外文关键词：Ulmus pumila sparse forest; Otingdag Sandy Land; tree detection; crown extraction; GF-2 data; automated extraction algorithm

摘要：The sparse Ulmus pumila L. woodland in the Otingdag Sandy Land of China is indispensable in maintaining the ecosystem stability of the desertified grasslands. Many studies of this region have focused on community structure and analysis of species composition, but without consideration of spatial distribution. Based on a combination of spectral and multiscale spatial variation features, we present a method for automated extraction of information on the U. pumila trees of the Otingdag Sandy Land using very high spatial resolution remote sensing imagery. In this method, feature images were constructed using fused 1-m spatial resolution GF-2 images through analysis of the characteristics of the natural geographical environment and the spatial distribution of the U. pumila trees. Then, a multiscale Laplace transform was performed on the feature images to generate multiscale Laplacian feature spaces. Next, local maxima and minima were obtained by iteration over the multiscale feature spaces. Finally, repeated values were removed and vector data (point data) were generated for automatic extraction of the spatial distribution and crown contours of the U. pumila trees. Results showed that the proposed method could overcome the lack of universality common to image classification methods. Validation indicated the accuracy of information extracted from U. pumila test data reached 82.7%. Further analysis determined the parameter values of the algorithm applicable to the study area. Extraction accuracy was improved considerably with a gradual increase of the Sigma parameter; however, the probability of missing data also increased markedly after the parameter reached a certain level. Therefore, we recommend the Sigma value of the algorithm be set to 90 (+/- 5). The proposed method could provide a reference for information extraction, spatial distribution mapping, and forest protection in relation to the U. pumila woodland of the Otingdag Sandy Land, which could also support improved ecological protection across much of northern China.