| Forest stock volume is a significant index in surveys and monitoring of forest resources.Using high-efficiency and high-accuracy method for estimating forest volume can reduce theground investigation work, which also obtain fast and accurate prediction of forest volume. Inthe process of forest stock volume estimation, the speed and precision of remote sensingimage classification has a direct influence on the estimation efficiency, even the reliability ofestimation results. Compared with the traditional remote sensing image classification softwareand methods, eCognition has been markedly improved in classification speed and accuracy,which has the object-oriented classification method.In the paper, firstly, the author uses eCognition to segment remote sensing image ofexperiment forest region, base on correlation coefficients. Secondly, the main influentialfactors of remote sensing and GIS of stock volume estimation by residual analysis has beenselected on the basis of realizing forest region remote sensing image, DEM and base mapsuperimposed display. Use stepwise regression analysis, ridge regression and partial leastsquares regression analysis to establish the volume estimation equation, by comparing theVolume of the experimental forest areas predicted residual sum of squares with the modelswhich established by three methods, and find out the best solution method for experimentalforest area volume estimation equation based on sample plot for the unit.Thirdly, the authorselects the optimal sample plot combination in experimental forest region which based onobserved value of forest volume, and establishes stock volume estimation model of sampleplot for the unit with regression analysis by SPSS. The Experiment shows that Partial leastsquares regression analysis to determine suitable experimental forest area volume estimationequation solver method is reliable, So we could estimate forest volume in the experimentalforest areas based on sample plot for the unit. |
PDF Full Text Request