| Forest resources were playing a very important role for adjusting and supporting on global ecological system,carbon cycle,energy conservation and human ecological environment.At the same time,it also had quite vital effect for the scientific research and monitoring on the growth and distribution of forest resources,for improving the production of forestry and sustainable economic and other benefits.Besides,the heat of remote sensing vegetation classification research has also been stimulated for the resolution of image that has improved constantly and emergence of new algorithm.The rotation forestis a kind of integrated classification algorithm based on feature transformation.The Rotation Forestalgorithm was used to classify forest vegetation,to some extent,which has possessed positive meaningin the development and research for forest resources.Taken Daxing'an Mountains of Heilongjiang Province as study area and based on Landsat8 OLI images data in 2013,1008 fixedplots data in 2010,sub-compartment distribution map in 2001,DEM and meteorological data in 2003(including the average precipitation and temperature)as the data source,three kinds of characteristic combination were acquired by increasing NDVI,NDBI,texture and topographic featureson the basis of the OLI bands(F1: OLI bands and NDVI,NDBI;F2: OLI bands,NDVI,NDBI and texture feature;F3: OLI bands,NDVI,NDBI,texture and topographic features),the rotating forest algorithm was used respectively to under the combination of forest vegetation classification,the highest feature combination of the classification accuracy was obtained,then MLC and SVM under the best feature combination were compared and analyzed.Later,on the basis of vegetation classification result of Rotation Forest classification,elevation,slope and aspect through DEM were extracted,average precipitation and average temperature figure were obtained which was taken on the meteorological data interpolation.Spatial distribution of various vegetation types and impact of the five factors on the spatial distribution of vegetation were analyzed by using the statistical analysis of spatial analysis tools.The results were showed that the highest accuracy of the feature groups was 86.85% by using the Rotation Forest algorithm and combining the OLI bands,texture and topographic features,respectively 10.14% and 2.99% than the accuracy of F1 ? F2 feature groups.Through comparing various classification method,the accuracy of Rotation Forest algorithm on forest vegetation had improved respectively 12.71% and 4.14% than the ones of MLC and SVM.Due to the better stability of Rotation Forest algorithm,it had relatively lower impact in the vegetation classification.Hence,the phenomenon of salt and pepper behaves the least,with the image being clearer and classification effect being better.Through the statistical analysis,In the Daxing'an Mountains,all kinds of vegetation coverage area increase firstly and then decrease with the increase of elevation.The elevation of 400-600 m range can be regarded as the most suitable range for vegetation growth.Besides,the slope has some impact on the environment for vegetation growth.In detail,all kinds of vegetation types focus on the flat and gentle slope.To be more specific,with the increase of slope,especially from the beginning of the steep slope,vegetation area fell sharply;the slope has little impact on the distribution of forest vegetation.In the four kinds of aspect,the vegetation distribution is relatively uniform,vegetation growth in the sunny and semi sunny slope are more than in shady and half shady slope and the tree species being more clutter;The precipitation in the Daxing'an Mountains concentrates between 600-700 mm.In this area,there are the most widely distributed range of vegetation.What's more,temperature is affected by terrain and latitude,increasing from northwest to southwest,mainly concentrated in about 1?.Especially between-2-1?,the area proportion of all sorts of vegetation types is larger. |
PDF Full Text Request