| This research,which is based on UAV remote sensing technology,carried out research work on the Dongtai Huanghai National Forest Park in Jiangsu Province as a target,and carried out several tree parameter extractions for 20 poplar plantation plots.In this study,a number of accuracy tests and regression models were fitted by combining the plot data and single-wood parameter data of the 2017 manual field survey.The main research results include the following aspects:(1)The drone height parameter setting plays an important role in image modeling:It is generally believed that the lower the UAV's altitude is,the higher the spatial resolution achieved by splice modeling.However,it has been found in practice that too low aerodynamic height can cause adjacent images to fail to match enough feature points,and thus cannot complete the splicing modeling work.(2)The drone overlap parameter setting plays an important role in modeling accuracy:By comparing the low-altitude drone image point cloud data of 18 plots,it was found that the number of point clouds generated by 95%overlapping degree compared to 90%overlapping degree has obviously increased,and the average lifting level reached117%.Because the Pix4D Capture IOS software limits the maximum overlap to 90%,this study proposes a double orbital flight method to achieve an increase in image overlap.(3)This study proposes a method for automatic extraction of canopy closure based on low-altitude drone DSM data:Based on the DSM data imaging mode and the characteristics of drone low-altitude photography,the rapid extraction of forest canopy environment was realized.Combining the manual survey data with the accuracy test results,the results show that the correlation coefficient R is 0.94,and the root mean square error RMSE is 0.02.The relative error RE%is 3.34%,and the Estimation accuracy EA is 96.66%.The above indicators perform well.The paired T test found that in the 95%confidence interval,the significance index obtained was higher than 0.05,and the acquired value was 0.583,indicating that the difference between the DSM extracted value and the actual measured value of UAV was not significant.This proves that the method of extracting canopy based on low altitude DSM data of UAV is of high precision and good stability,and can be an alternative to the traditional method of measuring canopy density.(4)The study demonstrated the feasibility of extracting single wood canopy by UAV remote sensing technology:Based on the drone DOM data,supplemented with the Envi object-oriented segmentation method and visual interpretation method,the extraction of 213poplar canopy was performed,and the accuracy of field survey data was combined.The results showed that:The correlation coefficient R of 0.89,root mean square error RMSE of 0.60,relative error RE%of 7.93%,and and the Estimation accuracy EA is 92.07%.Further analysis found that the other parameters of the model fitting were normal and the overall accuracy was good,indicating that the current method based on UAV remote sensing image extraction of single wood canopy has certain feasibility,but there is still room for improvement.(5)Establish Sanwu poplar single wood parameter model:Using the regression model,the establishment of poplar DBH and canopy model,diameter and canopy and tree height model,allows data to work between the inversion of individual tree parameters.The results showedthatthebestmodelofDBPandcanopyforpoplarwas D=0.123K~3-2.086K~2+15.582K-21.819,and the best model for DBH and canopy and tree height was D=1.956K+1.144H-11.028.Compared with the bivariate model of breast height and crown height and tree height,the accuracy of the DBH and canopy of the one-element model is greatly improved.R~2 is increased from 0.54 in the one-element model to 0.88 in the binary model. |
PDF Full Text Request