| Forest structure parameters is a very important parameter for forest growth and harvest, it is also an important factor in forest survey, operation and management. It can reflect the growth status of forest environment and forest growth condition directly. Grasp the stand structure parameters is pretty vital in order to timely analyze the status of forest resources and ecological environment change. Evergreen broad-leaved mountain is widely distributed in the southwest of Sichuan, which played a very important role in the ecological balance of the entire Yangtze River. However, there is a lack of the research of remote sensing images based on the region spatial structure parameters estimation, so via the high resolution remote sensing images to estimate the regional evergreen broad-leaved forest in the diameter at breast height, tree height and crown width has an important significance on the study of Southwest montane evergreen broadleaved forest ecologic processes.This paper takes Ya’an Yucheng town as the study area, utilized the ground survey data, the Pleiades remote sensing image geometric correction, atmospheric correction, supervised classification, extraction and remote sensing factors and geographical factors. The values of variable factors is filtered by the importance in the projection screen (VIP). Estimation models of the regional evergreen broad leaved forest diameter, breast height, tree height and crown width were established by partial least squares regression method. Based on the output of these models, the distributions results of evergreen broad-leaved forest in the diameter at breast height, tree height and crown width in the study area are accomplished. Finally, DBH, crown height and DBH, and crown width models were established by employing the estimation values; the accuracy of the models is evaluated by using ground measured data. The main conclusions are as follows:(1) The Pleiades image geometric correction, atmospheric correction and image cropping and processed by ENVI5.0 software, carries on the classification to the image with the "classification", combined with the ground check points on the precision of classification result inspection, image classification, the overall accuracy is 83.2%, achieved the precision requirements of remote sensing image classification. Combined with the characteristics of high resolution DEM and Pleiades images of the study area, the study area were extracted from the shadow of abundance, single band gray value, vegetation index, terrain extracting factor. Totally, there are 23 independent variables.(2) In 23 independent variables, there is a good correlation with the shadow of abundance (SF), environmental vegetation index (EVI) and canopy density (YBD) and diameter at breast height, tree height and crown diameter of stand. And the projection in the modeling of the important value of VIP is relatively high, that shadow abundance (SF), environmental vegetation index (EVI) and canopy density (YBD) can be used as an important factor of the remote sensing estimation of DBH, tree height and crown width; texture factor also has some factors and DBH, tree height and crown diameter have some correlation, but the correlation coefficient is not high; the single band, band ratio vegetation index, environment vegetation index and terrain factors and DBH, tree height and crown width of the correlation is not very significant, significant only individual factor and the DBH, tree height and crown diameter relationship(such as B2, B1/B2and the diameter at breast height DBH, tree height, elevation and NDVI1 and crown width), but the correlation coefficient is not high.(3) VIP for variable selection by projecting the importance value, select 8 variables to establish the remote sensing estimation of DBH by partial least squares regression method, Based on the test of the remaining samples, the overall accuracy precision of the estimation model is 83.4%, shadow abundance model (SF) of the maximum diameter of VIP. The final model expression after the standard coefficient inverse conversion of DBH is:D=-12.0022×SF-0.171831×YBD-0.002472×EVI+0.00415×ELEVATION-2.61318× (B1/B2)+2.62776×ENT-5.81345×ASM+3.50494×IDM+31.7432(4) VIP for variable selection by projecting the importance value, select 8variables to establish the remote sensing estimation of tree height by partial least squares regression method, shadow abundance model (SF) of the VIP value of the largest, Based on the test of the remaining samples, the overall accuracy precision of the estimation model is 84.7%, The final model expression after the standard coefficient inverse conversion of DBH is:H=-3.49087×SF-0.048134×YBD-0.0007111×EVI+0.571895×CON+0.790206xENT+ 0.000711×ELEVATION-1.185916×ASM+1.10686×IDM+14.32545(5) VIP for variable selection by projecting the importance value, select 9 variables to establish the estimation model of the crown by partial least squares regression method, based on the test of the remaining samples, the overall precision precision of estimation models with is 81.5%, the degree of canopy model (YBD) of the VIP value of the largest. The final model expression after the standard coefficient inverse conversion for the crown:C=7.28018×NDVI1-0.002901×EVI-0.9.95044×SF-6.44699×ASM+2.44982×CON+2. 95966×ENT+5.34899×IDM-0.241695×YBD-0.003809×ASPECT+43.7494(6) The model of the measured sample tree height and DBH and crown diameter estimates were established in the study area (D-H) model of tree height, diameter crown’ (D-C) model and the height and DBH, crown width (D-H-C) model, the model expressions are as follows:D-H model:D=2.9216×H-16.147;D-C model:D=0.7556×C-5.2241;D-H-C model:D=1.4971 xH+0.3872xC-11.2562;The overall accuracy of the three models is 82.4%,85.5% and 88.5%, respectively. It indicated that the fitting results of the three models are almost ideal, there is a certain enforceability significance in the practical applications. Estimation of D-H-C model is higher than that of D-H model and D-C model. |
PDF Full Text Request