Research On The Estimation Of Classified Forest Stock Volume Based On GIS And Remote Sensing Images

Forest stock volume as an important indicator of the mount of forest that is one of thebasic indicators of reflecting the total size of the forest resources and levels of a country orregion,is basis for planning deforestation and forest management. With the rapiddevelopment and application of GIS and remote sensing technology, it has become a hotforestry research topic to use high-resolution remote sensing images combined with limitedground survey plots in forest stock volume estimation. At present traditional multiple linearregression method is adopted by the domestic stock volume estimation based on GIS andremote sensing research, and results in decrease of estimation accuracy. This article willexamine five common algorithms in stock volume estimation, create and prefer classificationmodel to improve the accuracy and model applicability.The premise of stock volume estimation is that the classification of remote sensingimages can be effectively divided into corresponding type. First, we need preprocess theimages as much as possible to restore the target reflectance characteristics and the correctgeometric position. Then we need interpret information manually on woodland and establishinterpret standards. Finally, to make supervised classification of remote sensing images and tore-encode.In order to verify the effect of classification estimation and Box-Cox transformation, weuse the A-class inventory plots of Miyun County Seventh National forest resources, the sameperiod TM image data through geometric correction and30m resolution DEM data as basicdata, to classify and estimate forest volume of Miyun County in Beijing, and to correct thestock volume by Box-Cox transformation. The analysis results are as follows:Stock volume estimation by classification can effectively improve the estimationaccuracy and the ability of model forecast and classification modeling can be effectively usedinto stock volume estimation; Under the same conditions of factors and model sample plots,although the accuracy is slightly reduced after correcting the model by Box-Cox correction of stock volume, it can greatly improve the prediction ability and applicability of the model.Stock volume estimation needs regression diagnostics. When the abnormal plots are more andthey affect Gauss-Markov assumptions, we can introduce Box-Cox transformation.