Estimation Of Structural Parameters Of Plantation Forests Using UAV-based Oblique Photogrammetry

A comprehensive and accurate acquisition of structural parameters,such as tree height,breast diameter,crown width,and biomass,can provide a data basis for forest resource investigation and management.China has the largest area of plantation forests in the world,so the scientific management of planted forests is of great significance for efficient and high-quality forestry management and production.The traditional way of obtaining forest parameters is often through manual field surveys.The survey plots are difficult to lay out in a complex forest environment,and the data update frequency is low.It also consumes a lot of manpower and material resources,and the accuracy of the measurement data cannot meet the application requirements.With the rapid development of UAV near-earth remote sensing technology,oblique photogrammetry has been a new method of low-cost and high-efficiency measurement.In addition to obtaining three-dimensional structural information,it also has rich spectral and texture information,and has been widely used in City 3D modeling and smart tourism,but there are still few applications in forestry.Using UAV oblique photogrammetry technology to extract individual tree and stand parameters will have great research potentials and application perspectives.In order to explore the feasibility and key technology of extracting the structural parameters of plantation forests by UAV-based oblique photogrammetry,this paper takes Wangyedian Experimental Forest Farm of Karaqin Banner,Chifeng City,Inner Mongolia Autonomous Region as the research area,larch and Chinese pine plantations of different forest ages as the research objects.Totally 35 larch plots and 42 Chinese pine plots were set up and ground surveys were conducted,including 6 plots for individual tree parameters extraction.First,three-dimensional forest model was reconstructed from the original oblique photographs to generate dense point clouds,canopy height model(CHM)and digital orthophoto(DOM).Then at the individual tree scale,point cloud segmentation(PCS),CHM-based watershed segmentation(Watershed),and DOM-based object-oriented multi-resolution segmentation(MRS)algorithm were used for individual tree segmentation,tree height and crown area extraction;Based on the individual tree crown area extraction results,height metrics generated from individual tree point clouds,the individual tree biomass estimation model were constructed and the accuracy was evaluated with the survey data.At the stand scale,the vegetation index was constructed based on DOM generated from oblique photogrammetry data,and combined with the phenological characteristics of the vegetation for tree species classification of larch and Chinese pine;And then stepwise regression analysis were used to determine the best variable combinations for fitting the optimal biomass model with point cloud height metrics,vegetation index and density variables.The results are as follows:(1)The PCS method is suitable for larch individual tree segmentation with an average F of 0.94,extracting tree height of mature larch,mature and young Chinese pine with an average r RMSE of 3.9%,4.3%and 9.8%,respectively,and extracting the crown area of young Chinese pine with r RMSE of33.5%;The Watershed segmentation algorithm based on CHM is suitable for segmentation of Chinese pine with an average F of 0.89,tree height extraction of individual young larch trees with r RMSE of7.4%,and crown area extraction of young or mature larch,and mature Chinese pine with average r RMSE of 23.2%,25.7%and 20.9%,respectively.(2)The individual tree segmentation results were used to extract individual tree point clouds height metrics,and combined with the extracted crown area to fit individual tree above-ground biomass estimation models for larch and Chinese pine.The crown area contributes a lot and participates in the construction of each model.The relative accuracy of biomass estimation of mature larch and Chinese pine is greater than that of young larch and Chinese pine,but R~2 is greater than 0.84,and r RMSE is13.2%,16.0%,9.9%and 18.2%,respectively.Overall,the accuracy of the individual tree biomass estimation model for mature stands is higher than young stands.The individual tree point clouds height metrics and crown area can fit the individual tree biomass model with good results.(3)At the forest stand scale,the vegetation index and phenological characteristics are used to classify larch and Chinese pine forests.Based on the visible RGB bands of the orthoimages,the vegetation index VDVI was constructed.By setting the suitable VDVI thresholds separately,Chinese pine stand was identified using the orthoimage obtained in the defoliation period,and the vegetation and non-vegetation were classified using the image of leaf period.In the vegetation areas,the low vegetation was removed by setting the normalized point cloud height threshold to obtain the coverage of the plantation forests;Finally,the larch and Chinese pine stand was extracted by overlaying the plantation forest coverage area and the Chinese pine area.The user accuracy and the producer accuracy are both above 90%,indicating vegetation spectral characteristics combined with phenological characteristics can successfully classify larch and Chinese pine.(4)At the forest stand scale,an above-ground biomass model was established by the fusion of spectral characteristics of photographic images,height and density variables of point cloud extraction.Based on the point cloud height metrics as the independent variables,and introducing the vegetation index and point cloud density metrics,the accuracy of the fitted biomass model is significantly improved.The R~2of the three biomass models of larch is 0.81,0.86 and 0.93,respectively,and the three biomass models of Chinese pine is 0.77,0.91 and 0.93,respectively.The vegetation index VDVI and the density metric D5 have a larger contribution rate to the model.This indicates that the UAV-based oblique photography data can effectively estimate forest biomass.The research results show that the UAV-based oblique photography technology has a good application effect in structural parameters extraction and above-ground biomass models construction.Under different algorithms,the structural parameters of individual tree are effectively extracted,and the biomass models of single wood or forest stand are fitted with different feature variable combinations,and its accuracy can meet the basic needs of forestry investigation.The purpose of this study is to explore the automate process of extracting the structural parameters of plantation forests at the individual tree and stand scales by UAV-based oblique photography technology,and to provide a certain reference for reducing the cost of forestry resources inventory and improving the level of intensive forest management.