DOI: 10.1016/j.jag.2013.08.011
Scopus记录号: 2-s2.0-84897586103
论文题名: A comparison of selected classification algorithms for mappingbamboo patches in lower Gangetic plains using very high resolution WorldView 2 imagery
作者: Ghosh A ; , Joshi P ; K
刊名: International Journal of Applied Earth Observation and Geoinformation
ISSN: 15698432
出版年: 2014
卷: 26, 期: 1 起始页码: 298
结束页码: 311
语种: 英语
英文关键词: Bamboo mapping
; Feature selection
; GLCM texture
; Pixel and object based classification
; Random forest
; Support vector machine
; WorldView 2
Scopus关键词: accuracy assessment
; algorithm
; artificial intelligence
; bamboo
; classification
; image analysis
; land cover
; land use
; pixel
; satellite imagery
; tropical environment
; Gangetic Plain
; India
; West Bengal
英文摘要: Bamboo is used by different communities in India to develop indigenous products, maintain livelihoodand sustain life. Indian National Bamboo Mission focuses on evaluation, monitoring and development ofbamboo as an important plant resource. Knowledge of spatial distribution of bamboo therefore becomesnecessary in this context. The present study attempts to map bamboo patches using very high resolution(VHR) WorldView 2 (WV 2) imagery in parts of South 24 Parganas, West Bengal, India using both pixeland object-based approaches. A combined layer of pan-sharpened multi-spectral (MS) bands, first 3principal components (PC) of these bands and seven second order texture measures based Gray Level Co-occurrence Matrices (GLCM) of first three PC were used as input variables. For pixel-based image analysis(PBIA), recursive feature elimination (RFE) based feature selection was carried out to identify the mostimportant input variables. Results of the feature selection indicate that the 10 most important variablesinclude PC 1, PC 2 and their GLCM mean along with 6 MS bands. Three different sets of predictor variables(5 and 10 most important variables and all 32 variables) were classified with Support Vector Machine(SVM) and Random Forest (RF) algorithms. Producer accuracy of bamboo was found to be highest when10 most important variables selected from RFE were classified with SVM (82%). However object-basedimage analysis (OBIA) achieved higher classification accuracy than PBIA using the same 32 variables,but with less number of training samples. Using object-based SVM classifier, the producer accuracy ofbamboo reached 94%. The significance of this study is that the present framework is capable of accuratelyidentifying bamboo patches as well as detecting other tree species in a tropical region with heterogeneousland use land cover (LULC), which could further aid the mandate of National Bamboo Mission and relatedprograms. © 2013 Elsevier B.V. Citation statistics:
资源类型: 期刊论文
标识符: http://119.78.100.158/handle/2HF3EXSE/79772
Appears in Collections: 气候变化事实与影响
There are no files associated with this item.
作者单位: Department of Natural Resources, TERI University, 10 Institutional Area, Vasant Kunj, New Delhi 110070, India
Recommended Citation:
Ghosh A,, Joshi P,K. A comparison of selected classification algorithms for mappingbamboo patches in lower Gangetic plains using very high resolution WorldView 2 imagery[J]. International Journal of Applied Earth Observation and Geoinformation,2014-01-01,26(1)