The Research On Extracting Of Subcompartment Based On Remote Sensing

This research is mainly about the variation information of the subcompartment landtype.Find the changed information by using the data of ZY-1-02C image data,and the variationdetection and estimation combined with subcompartment data extracting.This paper focuse on the variation information of the type of the landscape situationbetween year2007and year2012using subcompartment data of year2007and ZY-1-02Cimages data, field survey data is used,with the data analysis application such as ArcGIS9.3、ERDAS9.2、ENVI4.7、Matlab7.0eCognition8.0and EXCEL,etc.Two methods are used toabstract the variation information,they are classify the images and compare the differences ortransform and use the the method of object-oriented classification.Firstly,some processes aredone such as geometrical correction,image clipping,radial correction,elimination cloud andshadow and image fusion.The main content and results as follows:(1)There are two methods to abstract the variation information for the remote imagesbetween year2007and year2012,by classify the images and then compare the differences or usethe method of object-oriented classification.The first method is classify the images of year2012first and then compare the different places based on field survey data of year2007,the other oneis classify the images by eCognition8.0and then compare the differernce places using fieldsurvey data.(2)Six ways of classify are used.They are parallelepiped method and mahalanobis distancemethod and maximum likehood method and SAM method and ANN,the last one is supportvector machine method.The results shows that the classification using the last method has thehighest overall accuracy.The forestry can be well recognized by computer,as the overall accuracycan reach93.37%.(3)Multi-scale segmentation are used first,this research try to use two scales to do thesegmentation:they are50and70.Then the fuzzy classification function is established to extractthe features.(4)The results show that during year2007and year2012,there are4kinds of situation in themajority according to the ZY-1-02C images and the field survey data of forest:non-forestland,forest land and other land which didn’t change during the time,non-forest land replace forestland or the forest land replace non-forest land.(5)Using the methods to classify the images and segment the image then extract the featuresdata.The results show that the non-changed areas are:the forest land has a area of1900.2734square kilometers,construction area is264.1502square kilometers, the waters area is17.2720square kilometers,the appropriate area is11.7561square kilometers, no-forest land area is6.0214square kilometers and the shrub land area is16.5338square kilometers.In the changed area,forest land converted to non-forest land area is small,10.6710square kilometers, and non-forest land converted to forest land area is34.6632square kilometers.(6)According to this research we find that it is effective to detecting the variationinformation used the data of02C ZY-1-02C.With the developing of the accuracy of remotesensing,we will get better result.