Vegetation Boundary Extraction From Remote Sensing Imagery

Vegetation is an important part of the ecological environment and geographicalenvironment, is one of the most intuitive information of remote sensing imagery. Vegetation inremote sensing imagery processing has a wide application, and an plays an important role in thethematic maps, urban planning, production estimating,vegetation growth monitoring, blaze anddrought monitoring, weather forecasting, geological structure research, military monitoringworks. Extracting the vegetation and it’s boundary information quickly and accurately in remotesensing imagery is particularly important. Special internal structure of the vegetation make itshowing certain characteristics in the imagery and can be effectively distinguished from otherobjects. This thesis analysis the various features of the vegetation in the remote sensing imagery,This thesis analysis the various features of the vegetation in the remote sensing imagery,thenThen respectively use vegetation index to extract grass from residents on the high resolutionimage of Quickbird and extract woodland from farmland on image of Quickbird andSPOT-5,and extract the boundary, finally verify the performance of the algorithm throughaccuracy assessment.This thesis research vegetation boundary extraction, main work are ordered as follows:1. Analysis plant spectral radiation characteristics, summarize the vegetation tone features,texture features and geometric features in remote sensing imagery, to establish a solid basis forvegetation boundary extraction.2. Analysis characteristics and field of application of these four kinds oftraditionalvegetation index, such us NDVI, DVI, RVI and SAVI. Aiming at traditional vegetation indexneed artificial threshold,put forward two kinds of fixed threshold vegetation index based on LBVtransform（NLVI and VBVI vegetation index）, take vegetation areas in residential areas asexperimental object, compare of these two index with four traditional index by experiments,summed up the advantages and disadvantages of the new algorithm and traditional algorithms.3. Extract typical vegetation region extraction method combination of texture features andspectral features.Obtaining texture imagery through single-band spectrum and spectruminterrelated GLCM,which fuse image texture features and spectral features, and then takeclassification use of the original multi-spectral imagery and texture imagery to obtain typicalvegetation area.This paper take woodland in farmland as object,Experimental results show thatthe algorithm can effectively extracted woodland area from farmland.4. Extracting The vegetation region boundary using morphological and boundary trackingmethod, evaluate boundary accuracy obtained by these two methods respectively by stats correct rate, error rate and missed detection rate, quantitative evaluation of the performance of these twomethods, and provide theory and data support for study in future.