Study On Identification And Tree-crown Extraction Of Populus Euphratica And Tamarix Ramosissima Based On High Resolution Remote Sensing Data

Taking QuickBird（QB） image, WorldView-1（WV） image, Unmanned Aerial Vehicle（UAV） imageas data source, Studying tree crown extraction effect of different method of Populus euphratica andTamarix ramosissima, analyzing the relationship between the crown diameters and groundwater depth, Themain conclusions are as following:1Tree species identification of Populus euphratica and Tamarix ramosissima based on high resolutionsatellite data（1）Using cross validation method to select support vector machine（SVM） parameter, determined thevalue of better penalty coefficient and kernel parameter. Different objects, different remote sensing data,different classification method whose parameters are different. Gradual transformation and observationmethod are used to determine the texture window. In the7×7windows, classification accuracy of QB datais the highest, up to86.14%; in the9×9windows, classification accuracy of WV data is the highest, up to84.17%. The results show that: Parameter selection determines the object classes recognition effect, treespecies identification accuracy of different parameters are different.（2）The overall classification mean accuracy of SVM method combined texture were84.01%,3.33%higher than single data source SVM method based on the QB data. The overall classification accuracy ofobject-oriented method of QB data and WV data is the highest, which reached87.79%、87.72%,respectively; Kappa coefficient reached0.84. In the lower reaches of Tarim River, object oriented method isthe best method for tree species identification based on high resolution remote sensing data, it caneffectively use the spectral, texture and spatial information of remote sensing data, solves the broken crownproblem of other methods caused by “same object with different spectrums” and “different objects sameimage”, improves tree species identification accuracy.2Extraction of crown diameters of Populus euphratica and Tamarix ramosissima based on high resolutionsatellite data（1）support vector machine method, object oriented method, maximum likelihood method was used toextract tree crown. Tree crown extraction accuracy of object oriented method is the highest in thosemethods, the crown diameters extraction accuracy of QB data and WV data is reached87.77%、89.41%,respectively. The results show that: tree crown extraction of a wide range of high resolution remote sensingdata can use object oriented method.（2）The accuracy of crown diameters extraction of Populus euphratica is higher than that of Tamarixramosissima; crown diameters extraction accuracy in higher density areas is lower than that of mediumdensity area and sparse region; the crown diameters extraction accuracy in the shallow zone of groundwateris higher than that of deep zone.3Crown diameters dynamics analysis of Populus euphratica and Tamarix ramosissima based on highresolution satellite dataThe paper extracted tree crown of Populus euphratica and Tamarix ramosissima based QB data（2004）and WV data（2011） in the same region of the study area, The results show that: crown diameters of2P0o0p4u~lu2s01e1u.Tphhrea ctilcoase ra tnhde dTisatmanacreix f rormam tohsei srsivimera, girnocurnedawseadt era dt epdtihff eisr emnot red isshtaanllcoews afrreoams, ththe eg reriavteerr thinecgrroowwnth aomf pclirtouwden idnitaemrveatle r5s~;7u syinegar sh iignh threes osalumtieo nr ergeimono,t et hsee nstsuindgy dparotav icdaen thbee bqausains tiftoart ivreemlyo toeb tsaeinns itnhgequantitative monitoring of forest resources and quantitative evaluation of ecological restoration.4Extraction of tree crown of Populus euphratica and Tamarix ramosissima based on UAV dataThe Object-oriented method and visual interpretation method were used to extract tree crowninformation of Populus euphratica and Tamarix ramosissima based on UAV data. The results show that:Single tree crown extraction accuracy of object-oriented method is86.72%,0.62%lower than the visualinterpretation method; The average crown extraction accuracy of object-oriented method is87.22%, higher1.86%than visual interpretation method; The plot canopy density estimation accuracy of object-orientedmethod is83.33%, higher3.19%than visual interpretation method. The crown extraction method ofobject-oriented overcomes the drawbacks of visual interpretation, which is slow and error varies fromperson to person. Object-oriented method is an effective method for large area of tree crown quantitative ofUAV data.