Research On Methods Of Forest Stand & Individual Tree Information Extraction Using UAV-based Aerial Photography

Forest survey aims to identify and implement the distribution and growth status of the national forest resources and provides data support for the sustainable development of forest resources.UAV aerial photography technology with high-resolution aerial images changes traditional ways of forest survey,which can extract a variety of forest survey parameters,thereby it reduces the intensity of field work and speeds up the development pace of digital forest survey.In this paper,fixed-wing and multi-rotor UAV are used as data acquisition platform,which are equipped with RGB band of ordinary digital camera as the sensor.Using the fixed-wing UAV collects the images data of a wide range of forest areas to extract the forest parameter information.And using the multi-rotor UAV collects the image data of the single-tree to extract the tree three-dimensional measurement information.The contents and conclusions are as follows:(1)High-resolution digital orthophotos and digital surface models are generated to extract three parameters of forest,which makes a single image into the entire survey area image.Using the two types of pixel feature statistics and object-oriented classification extracts the class information types of woodland,farmland,residential land,water,road and bare land.Effect of object-oriented classification extraction is best,which overall accuracy is 88.64%and Kappa coefficient is 0.85.Using two ways of object-oriented image segmentation and manual plotting vectorization extracts the crown width.The average precision of the extraction are 0.86 and 0.93 respectively,and the manual extraction is in higher precision.Using principal component regression method establishes inversion model of forest volume.The results show that the average relative error is 9.96%and the accuracy is 90.04%.(2)3D point cloud data of a target tree is reconstructed by the method of oblique photography.The 3D information of the tree measurement method is established to extract six factors.By selecting the measured point,the extraction precision of total tree height and first branch height are 96.28%and 95.61%respectively.The extraction precision of diameter at breast height(DBH)and upper diameter are 96.24%and 93.78%respectively.By using cutting method and projection method,the extraction accuracy of vertical canopy area are 96.28%and 98.24%respectively.The extraction accuracy of the crown width are 89.65%and 91.50%respectively.The accuracy of extraction crown surface area are 96.78%and 97.58%respectively.The accuracy of extraction crown volume are 94.29%and 96.14%respectively.In summary,the high resolution images obtained by fixed-wing UAV can realize the extraction of forest parameter information from the macroscopic scale,which can be applied to a wide range of forest mapping work.3D point cloud models obtained by multi-rotor UAV can realize the extraction of measurement factors of a target tree,which can be applied to the investigation and protection of ancient and famous trees.Therefore,the right types of UAV for forest information extraction can maximize the efficiency and accuracy with different purposes of forest surveys.