| Forest stock is an important basis to reflect the richness of forest resources and to measure the quality of forest ecological environment,and it is also one of the important indicators for the calculation of forest carbon reserves.After the flood disaster in 1998,the CPC Central Committee and the State Council made a major decision to implement the natural forest protection project.Since 2005,Xinjiang has imposed a total ban on logging,and the precious natural forest has been fundamentally protected.Picea schrenkiana var.tian schanica is the most important zonal forest vegetation in Tianshan Mountains,and it is the material main body of Tianshan and even Xinjiang forest ecosystem.Fast and accurate estimation of forest stock has always been a hot issue concerned by forestry workers.This study takes Picea schrenkiana var.tianschanica in Qiaxi National Forest Park of Gongliu Branch of Tianshan State-owned Forest Administration as the research object,using UAV images and World View-2 satellite images,on the basis of extracting the crown width and tree height information of Picea schrenkiana var.tianschanica,the volume estimation models based on"DBH-crown width model","DBH-tree height model"and"spectrum+texture"features were constructed respectively.The main results are as follows:In the aspect of crown information extraction of Picea schrenkiana var.tianschanica forest,when the multi-scale segmentation scale of UAV data was 74,the compactness factor was 0.5,and the shape factor was 0.3,the segmentation effect of Picea schrenkiana var.tian schanica crown was the best,and the average accuracy of extracting crown width was84.16%.Verified by the average accuracy method,the lowest accuracy of crown segmentation was 58.70%,the highest accuracy was 98.88%,and the overall accuracy was80.66%.In terms of tree height information extraction of Picea schrenkiana var.tianschanica forest,the use of UAV aerial images to establish a digital elevation model and digital surface model has shown that the average extraction accuracy of single tree height,based on canopy height model was 88.49%.The lowest extraction accuracy of average tree height of sample plot was 57.84%,the highest was 98.86%,and the overall accuracy was 88.13%.In terms of volume inversion of Picea schrenkiana var.tianschanica forest based on UAV images,the fitting degree of DBH-tree height model(R~2=0.89)was higher than that of DBH-crown width model(R~2=0.72),which reflected that the relationship between DBH and crown width was weaker than that between DBH and tree height.Moreover,at the sample level,the average accuracy of stock inversion based on"DBH-tree height model"(86.79%)was significantly higher than that based on"DBH-crown width model"(72.17%).Based on the inversion of Picea schrenkiana var.tianschanica forest volume based on World View-2 satellite images,with the participation of spectral information,vegetation index and texture factors,using random forest algorithm and taking 1500 as the number of regression trees,the regression model of sample plot stock and remote sensing factors was screened and established,and the degree of explanation was 81.27%.Through the applicability test and accuracy verification,the average accuracy of sample plot volume inversion can reach 82.49%.It is feasible to retrieve the stock of Picea schrenkiana var.tian schanica through UAV and satellite images.the inversion accuracy of Pi cea schrenk iana var.tianschanica volume inversion regression model based on high spatial resolution satellite data was not different from that of"DBH-tree height model"based on UAV images. |
PDF Full Text Request