Parameter Inversion Of Larch Pine Forest Based On High Resolution Remote Sensing Data

In recent years,using the strong correlation between the spectral features of remote sensing images and the stand parameters to inverse the stand parameters has been a hot spot of research.In previous studies,there are few studies on the inversion of high satellite data provided by China,and the research on the inversion of stand parameters for different resolution remote sensing images remains rare and to be explored.In order to further explore the domestic feasibility in remote sensing data in the estimation of forest parameters and applicability,and the influence of changing resolution characteristics of remote sensing inversion of forest parameters,this study used 5 different resolution domestic high resolution remote sensing image and terrain data,combined with the Larch ground survey sample data of study area,to study the larch volume and canopy density,and to explore the method of different resolution of domestic high satellite data inversion larch forest parameters,to achieve a high precision,large scale mapping of forest parameters,to provide a new technical reference for decision management of artificial forest.The main results are as follows:(1)The GF-2 remote sensing image data were used as representatives to discuss the influence of inter band registration errors on forest remote sensing information extraction.The correction error decreases first and then became larger when the number of control points increased.It was necessary to select a reasonable number of control points.There was a significant linear relationship between the area change and registration error in the multispectral image simulation registration error.In the application of the forestry industry,the registration error between multispectral bands of GF-2 image should be less than 0.3 pixels.It had a certain reference value for the domestic high resolution remote sensing image,and it could reduce the error influence caused by the pretreatment of high resolution remote sensing image.(2)In view of the few spectral bands of high resolution remote sensing images,it was difficult to distinguish the dominant tree species in the study area by using spectral differences.The differences in spectral characteristics of high resolution remote sensing images of different species were introduced,and different classification methods were used to distinguish the forest dominant tree species.In the extraction experiment of Larch remote sensing information,it was found that the classification accuracy of Larch was different from the remote sensing image of different phase.The total classification precision of the autumn image and the classification accuracy of Larch were higher than the summer images.The classification accuracy of the two kinds of time phase images was higher than the classification precision of single phase image,and the precision of Larch classification was the most up to 96.83%,the overall classification accuracy was up to 95.87%.(3)This study explored the variation of the best texture window.The size of the optimal window was related to the image resolution and the size of the plots.When the area of the window was close to the area of the sample,the size of the window was the best texture window.This provided a certain data reference for the selection of the optimal texture window of high resolution remote sensing image,and could reduce the amount of texture calculation.It was very practical.(4)The effect of resolution enhancement on the parameters of the deciduous forest based on image texture features was studied and analyzed.It was found that the enhancement of resolution could promote the inversion of the stand parameters by the texture features of remote sensing images,but it was not too high and the interference information of other non target types was increased when the resolution was too high.It would reduce the precision of the texture feature inversion model.In addition,when selecting the parameters of larch forest,it was better to choose high resolution remote sensing data with resolution over 8m as data source,and the accuracy of the model could reach over 0.6,This study selected ideal stand parameters estimation model for lm resolution remote sensing data fusion model based on GF-2.The study of image resolution change can provide a new method reference for improving the precision of forest stand parameter model to improve the texture feature of remote sensing image,and can be used for reference for the application of other high-resolution satellite image texture features in the model inversion.(5)The important influence of topographic factors on the inversion of the stand parameter model of high resolution remote sensing image was studied and discussed.It was found that the introduction of topographic information could better improve the precision of the model of the domestic high resolution remote sensing image to retrieve the parameters of the larch forest.With the increasing resolution of remote sensing images,the accuracy of the stand parameter model was increased gradually after the introduction of terrain information.The R2adj of the 8m resolution was 0.6,the accuracy of the 1m resolution GF-2 fusion image was the highest,the R value of the canopy and the forest volume model was 0.819 and 0.903,and the R2adj was 0.646 and 0.804.The applicability of remote sensing image inversion model in large scale and wide range was solved,and the gap of GF-2 remote sensing data in high precision quantitative inversion of different stand parameters was filled.It provides a new reference for the application of new domestic satellite data in the investigation and monitoring of forest resources.