Forest Mapping Using Multi-temporal ENVISAT ASAR Data

With the development of remote sensing technology at a very high speed,remotesensing are not only widely used in military,but also for civilian uses because of itsunique characteristics with timely high precision,high accuracy,all-weather andall-time capabilities.And by various ways(e.g.,radar data,aerial photographs,field data)to get data,it has a development tendency of more advanced,more intelligentized anddynamic monitoring insteading of static observation.When traditional visible light and infrared remote sensing can be affected by thelimitation of clouds,rain,snow,light conditions,synthetic aperture radar(SAR) canreplenish the defects of remote sensing with its all-weather and all-time capabilities toplay a tremendous role in forestry production.Whereas it is still immature in ourcountry,multi-temporal dual-polarized data are ultilized in this paper in classifing thedominant ground with forests.When land category and forests classification effect processed by thesingle-polarized,single band radar images is not good,the combination ofdual-polarized data obtained by ENVISAT ASAR AP and TM images is employed inthis paper to conduct the forest classification experiments.Among the various classic classification algorithms,no one is the best anduniversal,which is decided by remote sensing images,local site condition and thecomplexities of complex terrains and plants.More ground information can be obtainedfrom multi-temporal,multi-polarized radar images which can improve the accuracy offorest classification to achieve the aim of research.The conclusions are following:（1）Single-temporal ENVISAT ASAR images are pre-processed and classified withclassic classification methods(e.g.,Maximum likelihood,ANN,SVM).Compared withthe results,the precision of the one by SVM is better than the other two,which is notideal.（2）Classification with new composite images combined with multi-temporal,multi-polarized ENVISAT ASAR data band parameter can improve the precision offorest identity.（3）More ground information can be obtained from the union of multi-source data,soit can not only decrease the degree of obfuscation of forest and farmland but alsoimprove the precision of forest classification when classify the new images.